Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 1 year ago.
Improve this question
A little background here:
I know what a data warehouse is, more or less. I've read several dozen guides on data warehousing, I've played with SSAS, I know what a star schema and a dimension table and a fact table is, I know what ETL is and how to do it. This is not a "how" question or a request for tutorials.
My issue is that all of the material I've read on data warehousing seems to gloss over the rationale for building a data warehouse. They all figuratively, or in some cases literally start with the phrase "so you've decided to build a data warehouse..." Except I haven't made that decision yet.
So I'm hoping that SO members can point me to, or help come up with, some kind of semi-objective test. Something that I can adapt to a particular system and end up with either "yep, we need a data warehouse" or "no, the payoff today would be too small." I think that the specific questions I should be able to answer are:
At what point is building a data warehouse an option worth considering? In other words, what telltale signs, metrics, or other criteria should I be looking out for that might indicate that a standard transactional environment is no longer sufficient?
What are the alternatives to a full-on data warehouse? Denormalization in the transactional database and the bog-standard replicated "report server" are two that come to mind; are there any others I should explore before committing to the DW?
Why is a data warehouse better than said alternatives? If the answer is, "it depends", then what does it depend on?
When shouldn't I attempt to build a data warehouse? I'm skeptical of anything declared as a "best practice" irrespective of context. Surely there must be some scenarios where a DW is the wrong choice - what are they?
Are there any practical examples I could look at of systems that were improved by introducing a data warehouse? Something that would explain to me, end-to-end, what sorts of decisions or analysis they needed the warehouse for, how they decided what to put in it, and how the warehouse ended up fitting into the larger environment? I don't want a contrived "let's make a cube out of the AdventureWorks database" - the implementation is irrelevant to me, I'm interested in the specifications and designs and overall thought process that were involved.
I generally try not to ask multi-parters but I think that these are all very closely-related. I'm willing to accept any answer that addresses at least the first 4 questions, although the last would really help to crystallize this in my mind. Links are fine if somebody's already written about this, as long as they're reasonably concise and specific (link to Ralph Kimball's home page = not helpful).
Hope I've made the question clear - thanks in advance for your answers!
I'll see if I can do my best to answer your questions succinctly.
1.At what point is building a data warehouse an option worth considering?
In other words, what telltale signs,
metrics, or other criteria should I be
looking out for that might indicate
that a standard transactional
environment is no longer sufficient?
a. If you find that reporting and monitoring are impairing the performance of your production system and/or an offline data store.
b. If you find that getting answers to your business questions requires building a lot of complex SQL each time.
c. If you find that every time you make a change to your transactional schema, you have to go back and rework all of your reporting queries.
d. If you want to bring together data from multiple sources.
2.What are the alternatives to a full-on data warehouse?
Denormalization in the transactional
database and the bog-standard
replicated "report server" are two
that come to mind; are there any
others I should explore before
committing to the DW?
3.Why is a data warehouse better than said alternatives? If the answer is,
"it depends", then what does it depend
on?
I'll answer these together. I wouldn't think of a data warehouse as an all or nothing venture. It's simply a concise phrase that means "storing your data in a way that allows you to more easily and quickly answer business questions."
Transactional databases are designed to efficiently interface with applications. Data warehouses, data marts, operational data stores and reporting tables are built to efficiently interface with people, if that makes sense.
4.When shouldn't I attempt to build a data warehouse? I'm skeptical of
anything declared as a "best practice"
irrespective of context. Surely there
must be some scenarios where a DW is
the wrong choice - what are they?
Good question. If your transactional system provides you with sufficient insight into your business, you probably do not have a need for warehousing.
If you only have one source of data and performance is not a problem, you can probably gain insight from creation of simple reporting tables.
5.Are there any practical examples I could look at of systems that were
improved by introducing a data
warehouse? Something that would
explain to me, end-to-end, what sorts
of decisions or analysis they needed
the warehouse for, how they decided
what to put in it, and how the
warehouse ended up fitting into the
larger environment? I don't want a
contrived "let's make a cube out of
the AdventureWorks database" - the
implementation is irrelevant to me,
I'm interested in the specifications
and designs and overall thought
process that were involved.
That's a big question that would take far more space than I'm allotted here.
On this one, I can point you to a few places that might provide the insight you seek.
"Implementing A Data Warehouse: A Methodology that worked" by Bruce Ullrey is a book documenting one man's journey to building a data warehouse. It's not highly polished, which gives it more realism. It reads like a journal with lots of models and other visuals that illustrate his efforts pretty well.
"Business Intelligence Roadmap" by Larissa Moss. Standard fare. Walks you through the process of building a BI practice at a high level.
"The Profit Impact of Business Intelligence" by Steve Williams gives a number of case studies that show the value of building data warehouses.
The main purpose of a DW is to speed-up (simplify) reporting and analytic. It enables slicing and dicing of data in any way a business user can think of.
For a first step DW, you can simply implement a Kimball star schema and run SQL queries against it. If this proves to be still too slow, start thinking about pre-calculated aggregations (cubes).
The slicing and dicing of information against a DW is way simpler, than against a normalized DB. Replicated report server will improve performance, but will not simplify slicing and dicing. Also keep in mind that the DW belongs to business users, so it is up to them to come up with various slice/dice ideas at any time -- IT people should simply provide environment in which something like this is possible.
If you just run few reports from time-to-time on your operational system and are satisfied with performance, there is no need for DW.
All my experience is with systems where business users endlessly complain about slow reports and inability to write "complicated queries", while production people complain that the database gets bogged down due to reporting. In all cases a simple Kimball star and a report server with cache and snapshots were good enough.
You should consider building a data warehouse, when two of the following criteria match:
Huge amount of data
Many big complex selects (possibly compared to few inserts, updates, and deletes) that just take too long to execute (and are complicated to write)
Data from different systems needs to get combined
It's really the question what you consider a data warehouse. In many cases you can move gradually from OLTPs Systems with some reports to a full blown data warehouse, as long as you can stick to a relational database management system. First could be to build a first fact table, and keep using the normalized tables for dimension. Then adding more facts, more fact tables or dedicated dimension tables to the game. First in the same database (or one of the databases of the involved systems), possibly moving to a separate database later.
A full data warehouse (separate database, star schema) offers the best options for tuning select statements, apart from going to a specialized system. It is also cleanly decoupled from the OLTP system(s). Think schema design, but also resources like CPU, I/O and memory and organizational, like scheduling of new releases. Of course it is a lot of work which you possibly don't need.
It's in the answers above: just because you have a handfull of complex queries, doesn't mean you should build a DWH, same holds for the other criteria, if they come in isolation.
Can't offer much here, but the advice: go agile. The requirements for a DWH depend extremly on the possibilities the users see. There for requirements are likely to change. Automating tests with databases is a pain, but fooling around in a production system with no proper tests is worse.
At what point is building a data warehouse an option worth considering? In other words, what telltale signs, metrics, or other criteria should I be looking out for that might indicate that a standard transactional environment is no longer sufficient?
I'd recommend a data warehouse when you observed that performing reporting and analysis activities on the in the transactional data store was harmful to both.
What are the alternatives to a full-on data warehouse? Denormalization in the transactional database and the bog-standard replicated "report server" are two that come to mind; are there any others I should explore before committing to the DW?
I have nothing to offer here. I'd say that keeping the transactional and reporting databases seems sensible to me, regardless of whether you call it a warehouse or not. Data mining can be a very CPU intensive activity.
Why is a data warehouse better than said alternatives? If the answer is, "it depends", then what does it depend on?
I have nothing to offer here.
When shouldn't I attempt to build a data warehouse? I'm skeptical of anything declared as a "best practice" irrespective of context. Surely there must be some scenarios where a DW is the wrong choice - what are they?
I'd say that if you don't need to keep long history, aren't doing intensive analysis of the data, and your reporting needs are limited to an ad hoc query from time to time, then perhaps a data warehouse isn't necessary.
Are there any practical examples I could look at of systems that were improved by introducing a data warehouse? Something that would explain to me, end-to-end, what sorts of decisions or analysis they needed the warehouse for, how they decided what to put in it, and how the warehouse ended up fitting into the larger environment? I don't want a contrived "let's make a cube out of the AdventureWorks database" - the implementation is irrelevant to me, I'm interested in the specifications and designs and overall thought process that were involved.
My employers have all used data warehouses for many years prior to my arrival, so I can't speak to what things were like before I arrived.
From my experience, the first sign for starting to think about data warehousing is when you have (or are developing) a transactional database and the users start adding lots of reporting and data history requirements. Which is pretty much always. It's always easier to have a separate data warehouse or reporting database than to try to design a transactional system that handles the reporting needs that end users always have. Storing history (for business entities) in a transactional system adds complexity and bloats a database that should be as responsive as possible.
On the flip side, I've been in large companies where many groups created data warehouses because data of interest was spread across many systems and was therefore difficult to query. The problem was that each group created their own data warehouse because all the existing warehouses in the company did not have the right subset of information, or had a data model that was regarded as non-optimal or incorrect. This made the situation worse by creating even more disparate data systems that were hard to compare.
DW could be considered if, one is using a ‘Transactional System’ from a long period. Later, they realize that they need to perform some data mining, to determine different data patterns of the business. And finally, with the help of the determined data patterns, one wants to help the top management to take further decisions in the benefit of the company.
Following steps needs to be taken up for building up a data ware house:
An ETL platform and database needs to be decided for the database.
A reporting tool like SSRS, Tableau, etc. needs to be chosen for the visualization.
One may opt for the Data Analytical language like R, for further use.
Finally, all this will help in developing the data ware house and reporting tool.
"I think that why do some projects fail?"
There are five primary reasons:
lack of partnership between the IT department and business users;
incorrect data warehouse architecture;
not enough experienced people;
improper planning, such as failure to use a proven methodology and a plan to ensure that no details are omitted;
and depending on bleeding-edge technology.
Related
Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.
We don’t allow questions seeking recommendations for books, tools, software libraries, and more. You can edit the question so it can be answered with facts and citations.
Closed 6 years ago.
Improve this question
For a current project I'm creating a data model. Are there any sources where I can find "best practices" for a good data model? Good means flexible, efficient, with good performance, style, ... Some example questions would be "naming of columns", "what data should be normalized", or "which attributes should be exported into an own table". The source should be a book :-)
Personally I think you should read a book on performance tuning before beginning to model a database. The right design can make a world of difference. If you are not expert in performance tuning, you aren't qualified to design a database.
These books are Database specific, here is one for SQl Server.
http://www.amazon.com/Server-Performance-Tuning-Distilled-Experts/dp/1430219025/ref=sr_1_1?s=books&ie=UTF8&qid=1313603282&sr=1-1
Another book that you should read before starting to design is about antipatterns. Always good to know what you should avoid doing.
http://www.amazon.com/SQL-Antipatterns-Programming-Pragmatic-Programmers/dp/1934356557/ref=sr_1_1?s=books&ie=UTF8&qid=1313603622&sr=1-1
Do not get stuck in the trap of designing for flexibility. People use that as a way to get out of doing the work to design correctly and flexible databases almost always perform badly. If more than 5% of your database design depends on flexibility, you haven't modeled correctly in my opinion. All the worst COTS products I've had to work with were designed for flexibility first.
Any decent database book will discuss normalization. You can also find that information easily on the web. Be sure to actually create FK/PK relationships.
As far as naming columns, pick a standard and stick with it consistently. Consistency is more important than the actual standard. Don't name columns ID (see SQL antipatterns book). Use the same name and datatypes if columns are going to be in several different tables. What you are going for is to not have to use functions to do joins because of datatype mismatches.
Always remember that databases can (and will) be changed outside the application. Anything that is needed for data integrity must be in the database not the application code. The data will be there long after the application has been replaced.
The most important things for database design:
Thorough definition of the data needed (including correct datatypes)
and the relationships between pieces of data (including correct normalization)
data integrity
performance
security
consistency (of datatypes, naming standards etc.)
The best book I've read on the design of database systems was "An Introduction to Database Systems". Joe Celko's SQL for Smarties books are also worth reading.
Assuming you're building an application and not just a database, and assuming you're using an Object Oriented language, Applying UML and Patterns by Craig Larman has a good discussion on mapping databases to objects.
In terms of defining "good", in my experience "maintainable" is probably top of the list. Maintainability in database design means many things, such as sticking to conventions - I often recommend http://justinsomnia.org/2003/04/essential-database-naming-conventions-and-style/. Normalization is another obvious maintainability strategy. I often recommend being generous with column types - it's hard to change an application if you find out that postal codes in different countries are longer than in the US. I often recommend using views to abstract complex data relations away for less experienced developers.
A key thing with maintainability is the ability to test and deploy. It's worth reading up about Continuous Database Integration (http://www.codeproject.com/KB/architecture/Database_CI.aspx) - whilst not strictly associated with the design of the database schema, it's important context.
As for performance - I believe you should design for maintainability first, and only design for performance if you know you have a problem. Sometimes, you know in advance that performance will be a major problem - designing a database for Facebook (or Stack Exchange), designing a database with huge amounts of data (terabytes and up), or huge numbers of users. Most systems don't fall into that camp - so I recommend regular performance tests, with representative data, to find if you have a problem, and only tune when you can prove you have to. Many performance optimizations are at the expense of maintainability - denormalization, for instance.
Oh, and in general, avoid triggers and stored procedures if you can. That's just my opinion, though...
Even though it is not a book I recommend to read Query evaluation techniques for large databases. It gives a background on query processing which largely influences your schema design, especially for data intensive (e.g., analytical) workloads. It is less hands-on but I believe every database designer should read it at least once :-).
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 3 years ago.
Improve this question
http://sql.wikis.com/wc.dll?SQL~datawarehouse tells:
"Sid Adelman of Sid Adelman & Associates in a recent presentation observed that the Meta group estimates the cost of a single data warehouse implementation project runs around $3 million, and that is for a single, initial implementation, nowhere near the scope of providing integrated views for an entire Enterprise"
I am afraid $3 million does not tell anything to majority of ppl.
How does it relate to the cost of corresponding (by size and level of data processing of) OLTP database implementation?
Is it higher/lower? how many times?
Note that OLAP solutions are usually being implemented after the costs of DBMS were already made for OLTP solutions...
Why are the costs so elevated?
Update:
Let me reformulate the question:
Why are the OLAP solutions rather very rare in comparison with OLTP ones?
Does the laboriousness and costs of OLAP seem too prohibitive?
Nobody seems to doubt in the need and necessity to spend money on OLTP.
Though, from the logical point of view, it is not clear to me why it is not vice versa?
There are a lot of legacy data sources already accumulated even outside of DBMS...
Update2:
Reformulating the question again...
One can judge about professional and business activity in certain areas by activity (number and frequency) of forum posts questions, vacancies, etc.
OLTP related questions has 2 orders more frequency(number) of questions compared to OLAP ones in this SO site.
Why is it?
How does it relate to the cost of corresponding (by size and level of of data of) OLTP database implementation?
The question makes no sense. You might as well ask how OLAP implementations compare with buying a new Bentley Continental Automobile. Or ask how OLAP implementations compare with an SAP ERP implementation. Or ask how OLAP implementations compare with a vacation in the South of France.
There's nothing comparable between OLAP and OLTP except that they both use a database.
Is it higher/lower? how many times?
Yes. It can be higher, lower or the same. It depends on the scope of work, not the database architecture.
Why are the OLAP solutions rather very rare in comparison with OLTP ones?
According to whom? Everyone who implements a "reporting" system that is attached to their transactional system is doing OLAP. Many, many applications are in two parts: the core transactional part and some reporting add-ons.
I wouldn't call using Business Objects or Cognos "very rare".
Does the laboriousness and costs of OLAP seem too prohibitive?
OLAP costs depend on the scope of work. There's nothing inherently prohibitive. If you install BO to do some reporting, the cost is very small. If you create a large enterprise-wide warehouse, the cost is large.
Why are the costs so elevated?
Compared with what? Companies can spend $80M US implementing SAP. That's higher. But not comparable. Companies can spend nearly $0 (under $100K) implementing a free open-source component. That's lower. But not comparable.
Why is it? OLTP related questions has 2 orders more frequency(number) of questions compared to OLAP ones in this SO site.
That's obvious. OLAP is easy. OLTP is difficult.
Also C# is the most difficult programming language. Equally obvious.
Why is it? OLTP related questions has
2 orders more frequency(number) of
questions compared to OLAP ones in
this SO site.
Could it be that generally experienced database developers do OLAP while OLTP is often done by application programmers with little or no database knowledge? So there is less need to ask questions and the questions they have tend to be too complex for a forum.
Could there be fewer OLAP databases because they generally need one or more OLTP databases to pull data from and that OLAP implementations often consolidate data from many OLTP databases? Could it also be that small databases and databases on some subjects don't generally need data warehousing solutions at all? Generally there is no need to develop a data warehouse if the reporting needs perform adequately aginst the OLTP database, so it is generally only as the amount of data gets very large and the reporting gets very complex that these types of projects are initiated.
I can also assure you that our Enterprise OLTP databases cost considerably more than 3 million to develop. By comparison our OLAP databases were much cheaper. We have spent less that 5% of the development time on OLAP than on OLTP. Maybe even less than 1%.
There's very little point in comparing the costs of two vastly different sets of requirements. Anyway, there is no such thing as an "average" project. OLTP refers to one type of database workload whereas OLAP refers to a set of technologies used for decision support applications. They are literally incomparable. Nor do I understand on what you are basing your idea that OLAP is "rare". OLAP is only one type of solution, inevitably the universe of all other database solutions is much bigger than just OLAP.
EDIT:
Maybe you might rephrase the question this way: "Why is there less money spent on decision support database applications than there is on other types of database applications?". It depends how you define "decision support" of course. However, if you look at the proportion of most organisations' resources dedicated to "decision support" versus the resources used on doing other stuff then I think you will see the reason.
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 2 years ago.
Improve this question
There has been a lot of talk related to Cassandra lately.
Twitter, Digg, Facebook, etc all use it.
When does it make sense to:
use Cassandra,
not use Cassandra, and
use a RDMS instead of Cassandra.
There is nothing like a silver bullet, everything is built to solve specific problems and has its own pros and cons. It is up to you, what problem statement you have and what is the best fitting solution for that problem.
I will try to answer your questions one by one in the same order you asked them. Since Cassandra is based on the NoSQL family of databases, it's important you understand why use a NoSQL database before I answer your questions.
Why use NoSQL
In the case of RDBMS, making a choice is quite easy because all the databases like MySQL, Oracle, MS SQL, PostgreSQL in this category offer almost the same kind of solutions oriented toward ACID properties. When it comes to NoSQL, the decision becomes difficult because every NoSQL database offers different solutions and you have to understand which one is best suited for your app/system requirements. For example, MongoDB is fit for use cases where your system demands a schema-less document store. HBase might be fit for search engines, analyzing log data, or any place where scanning huge, two-dimensional join-less tables is a requirement. Redis is built to provide In-Memory search for varieties of data structures like trees, queues, linked lists, etc and can be a good fit for making real-time leaderboards, pub-sub kind of system. Similarly there are other databases in this category (Including Cassandra) which are fit for different problem statements. Now lets move to the original questions, and answer them one by one.
When to use Cassandra
Being a part of the NoSQL family, Cassandra offers a solution for problems where one of your requirements is to have a very heavy write system and you want to have a quite responsive reporting system on top of that stored data. Consider the use case of Web analytics where log data is stored for each request and you want to built an analytical platform around it to count hits per hour, by browser, by IP, etc in a real time manner. You can refer to this blog post to understand more about the use cases where Cassandra fits in.
When to Use a RDMS instead of Cassandra
Cassandra is based on a NoSQL database and does not provide ACID and relational data properties. If you have a strong requirement for ACID properties (for example Financial data), Cassandra would not be a fit in that case. Obviously, you can make a workaround for that, however you will end up writing lots of application code to simulate ACID properties and will lose on time to market badly. Also managing that kind of system with Cassandra would be complex and tedious for you.
When not to use Cassandra
I don't think it needs to be answered if the above explanation makes sense.
When evaluating distributed data systems, you have to consider the CAP theorem - you can pick two of the following: consistency, availability, and partition tolerance.
Cassandra is an available, partition-tolerant system that supports eventual consistency. For more information see this blog post I wrote: Visual Guide to NoSQL Systems.
Cassandra is the answer to a particular problem: What do you do when you have so much data that it does not fit on one server ? How do you store all your data on many servers and do not break your bank account and not make your developers insane ? Facebook gets 4 Terabyte of new compressed data EVERY DAY. And this number most likely will grow more than twice within a year.
If you do not have this much data or if you have millions to pay for Enterprise Oracle/DB2 cluster installation and specialists required to set it up and maintain it, then you are fine with SQL database.
However Facebook no longer uses cassandra and now uses MySQL almost exclusively moving the partitioning up in the application stack for faster performance and better control.
The general idea of NoSQL is that you should use whichever data store is the best fit for your application. If you have a table of financial data, use SQL. If you have objects that would require complex/slow queries to map to a relational schema, use an object or key/value store.
Of course just about any real world problem you run into is somewhere in between those two extremes and neither solution will be perfect. You need to consider the capabilities of each store and the consequences of using one over the other, which will be very much specific to the problem you are trying to solve.
Besides the answers given above about when to use and when not to use Cassandra, if you do decide to use Cassandra you may want to consider not using Cassandra itself, but one of the its many cousins out there.
Some answers above already pointed to various "NoSQL" systems which share many properties with Cassandra, with some small or large differences, and may be better than Cassandra itself for your specific needs.
Additionally, recently (several years after this question was originally asked), a Cassandra clone called Scylla (see https://en.wikipedia.org/wiki/Scylla_(database)) was released. Scylla is an open-source re-implementation of Cassandra in C++, which claims to have significantly higher throughput and lower latencies than the original Java Cassandra, while being mostly compatible with it (in features, APIs, and file formats). So if you're already considering Cassandra, you may want to consider Scylla as well.
I will focus here on some of the important aspects which can help you to decide if you really need Cassandra. The list is not exhaustive, just some of the points which I have at top of my mind-
Don't consider Cassandra as the first choice when you have a strict requirement on the relationship (across your dataset).
Cassandra by default is AP system (of CAP). But, it supports tunable consistency which means it can be configured to support as CP as well. So don't ignore it just because you read somewhere that it's AP and you are looking for CP systems. Cassandra is more accurately termed “tuneably consistent,” which means it allows you to easily decide the level of consistency you require, in balance with the level of availability.
Don't use Cassandra if your scale is not much or if you can deal with a non-distributed DB.
Think harder if your team thinks that all your problems will be solved if you use distributed DBs like Cassandra. To start with these DBs is very simple as it comes with many defaults but optimizing and mastering it for solving a specific problem would require a good (if not a lot) amount of engineering effort.
Cassandra is column-oriented but at the same time each row also has a unique key. So, it might be helpful to think of it as an indexed, row-oriented store. You can even use it as a document store.
Cassandra doesn't force you to define the fields beforehand. So, if you are in a startup mode or your features are evolving (as in agile) - Cassandra embraces it. So better, first think about queries and then think about data to answer them.
Cassandra is optimized for really high throughput on writes. If your use case is read-heavy (like cache) then Cassandra might not be an ideal choice.
Right. It makes sense to use Cassandra when you have a huge amount of data, a huge number of queries but very little variety of queries. Cassandra basically works by partitioning and replicating. If all your queries will be based on the same partition key, Cassandra is your best bet. If you get a query on an attribute that is not the partition key, Cassandra allows you to replicate the whole data with a new partition key. So now you have 2 replicas of the same data with 2 different partition keys.
Which brings me to your next question. When not to use Cassandra. As I mentioned, Cassandra scales by replicating the complete database for every new partitioning key. But you can't keep making new copies again and again. So when you have a high variety in queries i.e. each query has a different column in the where clause, Cassandra is not a good option.
Now for the third question. The whole point of using RDBMS is when you want the ACID properties. If you are building something like a payment service and want each transaction to be isolated, each transaction to either complete or not happen at all, changes to be persistent despite system failure, and the money to be consistent across bank accounts before and after the transaction completes, an RDBMS is the only option that will help you achieve this.
This article actually explains the whole thing, especially when to use Cassandra or not (as opposed to some other NoSQL option) part of the question -> Choosing the best Database. Do check it out.
EDIT: To answer the question in the comments by proximab, when we think of banking systems we immidiately think "ACID is the best solution". But even banking systems are made up of several subsystems that might not even be dealing with any transaction related data like account holder's personal information, account statements, credit card details, credit histories, etc.
All of this information needs to be stored in some database or the another. Now if you store the account related information like account balance, that is something that needs to be consistent at all times. For example, if you try to send money from account A to account B, then the money that disappears from account A should instantaneousy show up in account B, and it cannot be present in both accounts at the same time. This system cannot be inconsistant at any point. This is where ACID is of utmost importance.
On the other hand if you are saving credit card details or credit histories, that should not get into the wrong hands, then you need something that allows access only to authorised users. That I believe is supported by Cassandra. That said, data like credit history and credit card transactions, I think that is an ever increasing data. Also there is only so much yo can query on this data i.e. it has a very finite number of queries. These two conditions make Cassandra a perfect solution.
Talking with someone in the midst of deploying Cassandra, it doesn't handle the many-to-many well. They are doing a hack job to do their initial testing. I spoke with a Cassandra consultant about this and he said he wouldn't recommend it if you had this problem set.
You should ask your self the following questions:
(Volume, Velocity) Will you be writing and reading TONS of information , so much information that no one computer could handle the writes.
(Global) Will you need this writing and reading capability around the world so that the writes in one part of the world are accessible in another part of the world?
(Reliability) Do you need this database to be up and running all the time and never go down regardless of which Cloud, which country, whether it's VM , Container, or Bare metal?
(Scale-ability) Do you need this database to be able to continue to grow easily and scale linearly
(Consistency) Do you need TUNABLE consistency where some writes can happen asynchronously where as others need to be certified?
(Skill) Are you willing to do what it takes to learn this technology and the data modeling that goes with creating a globally distributed database that can be fast for everyone, everywhere?
If for any of these questions you thought "maybe" or "no," you should use something else. If you had "hell yes" as an answer to all of them, then you should use Cassandra.
Use RDBMS when you can do everything on one box. It's probably easier than most and anyone can work with it.
Heavy single query vs. gazillion light query load is another point to consider, in addition to other answers here. It's inherently harder to automatically optimize a single query in a NoSql-style DB. I've used MongoDB and ran into performance issues when trying to calculate a complex query. I haven't used Cassandra but I expect it to have the same issue.
On the other hand, if your load is expected to be that of very many small queries, and you want to be able to easily scale out, you could take advantage of eventual consistency that is offered by most NoSql DBs. Note that eventual consistency is not really a feature of a non-relational data model, but it is much easier to implement and to set up in a NoSql-based system.
For a single, very heavy query, any modern RDBMS engine can do a decent job parallelizing parts of the query and take advantage of as much CPU and memory you throw at it (on a single machine). NoSql databases don't have enough information about the structure of the data to be able to make assumptions that will allow truly intelligent parallelization of a big query. They do allow you to easily scale out more servers (or cores) but once the query hits a complexity level you are basically forced to split it apart manually to parts that the NoSql engine knows how to deal with intelligently.
In my experience with MongoDB, in the end because of the complexity of the query there wasn't much Mongo could do to optimize it and run parts of it on multiple data. Mongo parallelizes multiple queries but isn't so good at optimizing a single one.
Let's read some real world cases:
http://planetcassandra.org/apache-cassandra-use-cases/
In this article: http://planetcassandra.org/blog/post/agentis-energy-stores-over-15-billion-records-of-time-series-usage-data-in-apache-cassandra
They elaborated the reason why they didn't choose MySql is because db synchronization is too slow.
(Also due to 2-phrase commit, FK, PK)
Cassandra is based on Amazon Dynamo paper
Features:
Stability
High availability
Backup performs well
Read and Write is better than HBase, (BigTable clone in java).
wiki http://en.wikipedia.org/wiki/Apache_Cassandra
Their Conclusion is:
We looked at HBase, Dynamo, Mongo and Cassandra.
Cassandra was simply the best storage solution for the majority of our data.
As of 2018,
I would recommend using ScyllaDB to replace classic cassandra, if you need back support.
Postgres kv plugin is also quick than cassandra. How ever won't have multi-instance scalability.
another situation that makes the choice easier is when you want to use aggregate function like sum, min, max, etcetera and complex queries (like in the financial system mentioned above) then a relational database is probably more convenient then a nosql database since both are not possible on a nosql databse unless you use really a lot of Inverted indexes. When you do use nosql you would have to do the aggregate functions in code or store them seperatly in its own columnfamily but this makes it all quite complex and reduces the performance that you gained by using nosql.
Cassandra is a good choice if:
You don't require the ACID properties from your DB.
There would be massive and huge number of writes on the DB.
There is a requirement to integrate with Big Data, Hadoop, Hive and Spark.
There is a need of real time data analytics and report generations.
There is a requirement of impressive fault tolerant mechanism.
There is a requirement of homogenous system.
There is a requirement of lots of customisation for tuning.
If you need a fully consistent database with SQL semantics, Cassandra is NOT the solution for you. Cassandra supports key-value lookups. It does not support SQL queries. Data in Cassandra is "eventually consistent". Concurrent lookups of data may be inconsistent, but eventually lookups are consistent.
If you need strict semantics and need support for SQL queries, choose another solution such as MySQL, PostGres, or combine use of Cassandra with Solr.
Apache cassandra is a distributed database for managing large amounts of structured data across many commodity servers, while providing highly available service and no single point of failure.
The archichecture is purely based on the cap theorem, which is availability , and partition tolerance, and interestingly eventual consistently.
Dont Use it, if your not storing volumes of data across racks of clusters,
Dont use if you are not storing Time series data,
Dont Use if you not patitioning your servers,
Dont use if you require strong Consistency.
Mongodb has very powerful aggregate functions and an expressive aggregate framework. It has many of the features developers are accustomed to using from the relational database world. It's document data/storage structure allows for more complex data models than Cassandra, for example.
All this comes with trade-offs of course. So when you select your database (NoSQL, NewSQL, or RDBMS) look at what problem you are trying to solve and at your scalability needs. No one database does it all.
According to DataStax, Cassandra is not the best use case when there is a need for
1- High end hardware devices.
2- ACID compliant with no roll back (bank transaction)
It does not support complete transaction management across the
tables.
Secondary Index not supported.
Have to rely on Elastic search /Solr for Secondary index and the custom sync component has to be written.
Not ACID compliant system.
Query support is limited.
As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extended discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance.
Closed 9 years ago.
Whether we like it or not, many if not most of us developers either regularly work with databases or may have to work with one someday. And considering the amount of misuse and abuse in the wild, and the volume of database-related questions that come up every day, it's fair to say that there are certain concepts that developers should know - even if they don't design or work with databases today.
What is one important concept that developers and other software professionals ought to know about databases?
The very first thing developers should know about databases is this: what are databases for? Not how do they work, nor how do you build one, nor even how do you write code to retrieve or update the data in a database. But what are they for?
Unfortunately, the answer to this one is a moving target. In the heydey of databases, the 1970s through the early 1990s, databases were for the sharing of data. If you were using a database, and you weren't sharing data you were either involved in an academic project or you were wasting resources, including yourself. Setting up a database and taming a DBMS were such monumental tasks that the payback, in terms of data exploited multiple times, had to be huge to match the investment.
Over the last 15 years, databases have come to be used for storing the persistent data associated with just one application. Building a database for MySQL, or Access, or SQL Server has become so routine that databases have become almost a routine part of an ordinary application. Sometimes, that initial limited mission gets pushed upward by mission creep, as the real value of the data becomes apparent. Unfortunately, databases that were designed with a single purpose in mind often fail dramatically when they begin to be pushed into a role that's enterprise wide and mission critical.
The second thing developers need to learn about databases is the whole data centric view of the world. The data centric world view is more different from the process centric world view than anything most developers have ever learned. Compared to this gap, the gap between structured programming and object oriented programming is relatively small.
The third thing developers need to learn, at least in an overview, is data modeling, including conceptual data modeling, logical data modeling, and physical data modeling.
Conceptual data modeling is really requirements analysis from a data centric point of view.
Logical data modeling is generally the application of a specific data model to the requirements discovered in conceptual data modeling. The relational model is used far more than any other specific model, and developers need to learn the relational model for sure. Designing a powerful and relevant relational model for a nontrivial requirement is not a trivial task. You can't build good SQL tables if you misunderstand the relational model.
Physical data modeling is generally DBMS specific, and doesn't need to be learned in much detail, unless the developer is also the database builder or the DBA. What developers do need to understand is the extent to which physical database design can be separated from logical database design, and the extent to which producing a high speed database can be accomplished just by tweaking the physical design.
The next thing developers need to learn is that while speed (performance) is important, other measures of design goodness are even more important, such as the ability to revise and extend the scope of the database down the road, or simplicity of programming.
Finally, anybody who messes with databases needs to understand that the value of data often outlasts the system that captured it.
Whew!
Good question. The following are some thoughts in no particular order:
Normalization, to at least the second normal form, is essential.
Referential integrity is also essential, with proper cascading delete and update considerations.
Good and proper use of check constraints. Let the database do as much work as possible.
Don't scatter business logic in both the database and middle tier code. Pick one or the other, preferably in middle tier code.
Decide on a consistent approach for primary keys and clustered keys.
Don't over index. Choose your indexes wisely.
Consistent table and column naming. Pick a standard and stick to it.
Limit the number of columns in the database that will accept null values.
Don't get carried away with triggers. They have their use but can complicate things in a hurry.
Be careful with UDFs. They are great but can cause performance problems when you're not aware how often they might get called in a query.
Get Celko's book on database design. The man is arrogant but knows his stuff.
First, developers need to understand that there is something to know about databases. They're not just magic devices where you put in the SQL and get out result sets, but rather very complicated pieces of software with their own logic and quirks.
Second, that there are different database setups for different purposes. You do not want a developer making historical reports off an on-line transactional database if there's a data warehouse available.
Third, developers need to understand basic SQL, including joins.
Past this, it depends on how closely the developers are involved. I've worked in jobs where I was developer and de facto DBA, where the DBAs were just down the aisle, and where the DBAs are off in their own area. (I dislike the third.) Assuming the developers are involved in database design:
They need to understand basic normalization, at least the first three normal forms. Anything beyond that, get a DBA. For those with any experience with US courtrooms (and random television shows count here), there's the mnemonic "Depend on the key, the whole key, and nothing but the key, so help you Codd."
They need to have a clue about indexes, by which I mean they should have some idea what indexes they need and how they're likely to affect performance. This means not having useless indices, but not being afraid to add them to assist queries. Anything further (like the balance) should be left for the DBA.
They need to understand the need for data integrity, and be able to point to where they're verifying the data and what they're doing if they find problems. This doesn't have to be in the database (where it will be difficult to issue a meaningful error message for the user), but has to be somewhere.
They should have the basic knowledge of how to get a plan, and how to read it in general (at least enough to tell whether the algorithms are efficient or not).
They should know vaguely what a trigger is, what a view is, and that it's possible to partition pieces of databases. They don't need any sort of details, but they need to know to ask the DBA about these things.
They should of course know not to meddle with production data, or production code, or anything like that, and they should know that all source code goes into a VCS.
I've doubtless forgotten something, but the average developer need not be a DBA, provided there is a real DBA at hand.
Basic Indexing
I'm always shocked to see a table or an entire database with no indexes, or arbitrary/useless indexes. Even if you're not designing the database and just have to write some queries, it's still vital to understand, at a minimum:
What's indexed in your database and what's not:
The difference between types of scans, how they're chosen, and how the way you write a query can influence that choice;
The concept of coverage (why you shouldn't just write SELECT *);
The difference between a clustered and non-clustered index;
Why more/bigger indexes are not necessarily better;
Why you should try to avoid wrapping filter columns in functions.
Designers should also be aware of common index anti-patterns, for example:
The Access anti-pattern (indexing every column, one by one)
The Catch-All anti-pattern (one massive index on all or most columns, apparently created under the mistaken impression that it would speed up every conceivable query involving any of those columns).
The quality of a database's indexing - and whether or not you take advantage of it with the queries you write - accounts for by far the most significant chunk of performance. 9 out of 10 questions posted on SO and other forums complaining about poor performance invariably turn out to be due to poor indexing or a non-sargable expression.
Normalization
It always depresses me to see somebody struggling to write an excessively complicated query that would have been completely straightforward with a normalized design ("Show me total sales per region.").
If you understand this at the outset and design accordingly, you'll save yourself a lot of pain later. It's easy to denormalize for performance after you've normalized; it's not so easy to normalize a database that wasn't designed that way from the start.
At the very least, you should know what 3NF is and how to get there. With most transactional databases, this is a very good balance between making queries easy to write and maintaining good performance.
How Indexes Work
It's probably not the most important, but for sure the most underestimated topic.
The problem with indexing is that SQL tutorials usually don't mention them at all and that all the toy examples work without any index.
Even more experienced developers can write fairly good (and complex) SQL without knowing more about indexes than "An index makes the query fast".
That's because SQL databases do a very good job working as black-box:
Tell me what you need (gimme SQL), I'll take care of it.
And that works perfectly to retrieve the correct results. The author of the SQL doesn't need to know what the system is doing behind the scenes--until everything becomes sooo slooooow.....
That's when indexing becomes a topic. But that's usually very late and somebody (some company?) is already suffering from a real problem.
That's why I believe indexing is the No. 1 topic not to forget when working with databases. Unfortunately, it is very easy to forget it.
Disclaimer
The arguments are borrowed from the preface of my free eBook "Use The Index, Luke". I am spending quite a lot of my time explaining how indexes work and how to use them properly.
I just want to point out an observation - that is that it seems that the majority of responses assume database is interchangeable with relational databases. There are also object databases, flat file databases. It is important to asses the needs of the of the software project at hand. From a programmer perspective the database decision can be delayed until later. Data modeling on the other hand can be achieved early on and lead to much success.
I think data modeling is a key component and is a relatively old concept yet it is one that has been forgotten by many in the software industry. Data modeling, especially conceptual modeling, can reveal the functional behavior of a system and can be relied on as a road map for development.
On the other hand, the type of database required can be determined based on many different factors to include environment, user volume, and available local hardware such as harddrive space.
Avoiding SQL injection and how to secure your database
Every developer should know that this is false: "Profiling a database operation is completely different from profiling code."
There is a clear Big-O in the traditional sense. When you do an EXPLAIN PLAN (or the equivalent) you're seeing the algorithm. Some algorithms involve nested loops and are O( n ^ 2 ). Other algorithms involve B-tree lookups and are O( n log n ).
This is very, very serious. It's central to understanding why indexes matter. It's central to understanding the speed-normalization-denormalization tradeoffs. It's central to understanding why a data warehouse uses a star-schema which is not normalized for transactional updates.
If you're unclear on the algorithm being used do the following. Stop. Explain the Query Execution plan. Adjust indexes accordingly.
Also, the corollary: More Indexes are Not Better.
Sometimes an index focused on one operation will slow other operations down. Depending on the ratio of the two operations, adding an index may have good effects, no overall impact, or be detrimental to overall performance.
I think every developer should understand that databases require a different paradigm.
When writing a query to get at your data, a set-based approach is needed. Many people with an interative background struggle with this. And yet, when they embrace it, they can achieve far better results, even though the solution may not be the one that first presented itself in their iterative-focussed minds.
Excellent question. Let's see, first no one should consider querying a datbase who does not thoroughly understand joins. That's like driving a car without knowing where the steering wheel and brakes are. You also need to know datatypes and how to choose the best one.
Another thing that developers should understand is that there are three things you should have in mind when designing a database:
Data integrity - if the data can't be relied on you essentially have no data - this means do not put required logic in the application as many other sources may touch the database. Constraints, foreign keys and sometimes triggers are necessary to data integrity. Don't fail to use them because you don't like them or don't want to be bothered to understand them.
Performance - it is very hard to refactor a poorly performing database and performance should be considered from the start. There are many ways to do the same query and some are known to be faster almost always, it is short-sighted not to learn and use these ways. Read some books on performance tuning before designing queries or database structures.
Security - this data is the life-blood of your company, it also frequently contains personal information that can be stolen. Learn to protect your data from SQL injection attacks and fraud and identity theft.
When querying a database, it is easy to get the wrong answer. Make sure you understand your data model thoroughly. Remember often actual decisions are made based on the data your query returns. When it is wrong, the wrong business decisions are made. You can kill a company from bad queries or loose a big customer. Data has meaning, developers often seem to forget that.
Data almost never goes away, think in terms of storing data over time instead of just how to get it in today. That database that worked fine when it had a hundred thousand records, may not be so nice in ten years. Applications rarely last as long as data. This is one reason why designing for performance is critical.
Your database will probaly need fields that the application doesn't need to see. Things like GUIDs for replication, date inserted fields. etc. You also may need to store history of changes and who made them when and be able to restore bad changes from this storehouse. Think about how you intend to do this before you come ask a web site how to fix the problem where you forgot to put a where clause on an update and updated the whole table.
Never develop in a newer version of a database than the production version. Never, never, never develop directly against a production database.
If you don't have a database administrator, make sure someone is making backups and knows how to restore them and has tested restoring them.
Database code is code, there is no excuse for not keeping it in source control just like the rest of your code.
Evolutionary Database Design. http://martinfowler.com/articles/evodb.html
These agile methodologies make database change process manageable, predictable and testable.
Developers should know, what it takes to refactor a production database in terms of version control, continious integration and automated testing.
Evolutionary Database Design process has administrative aspects, for example a column is to be dropped after some life time period in all databases of this codebase.
At least know, that Database Refactoring concept and methodologies exist.
http://www.agiledata.org/essays/databaseRefactoringCatalog.html
Classification and process description makes it possible to implement tooling for these refactorings too.
About the following comment to Walter M.'s answer:
"Very well written! And the historical perspective is great for people who weren't doing database work at that time (i.e. me)".
The historical perspective is in a certain sense absolutely crucial. "Those who forget history, are doomed to repeat it.". Cfr XML repeating the hierarchical mistakes of the past, graph databases repeating the network mistakes of the past, OO systems forcing the hierarchical model upon users while everybody with even just a tenth of a brain should know that the hierarchical model is not suitable for general-purpose representation of the real world, etcetera, etcetera.
As for the question itself:
Every database developer should know that "Relational" is not equal to "SQL". Then they would understand why they are being let down so abysmally by the DBMS vendors, and why they should be telling those same vendors to come up with better stuff (e.g. DBMS's that are truly relational) if they want to go on sucking hilarious amounts of money out of their customers for such crappy software).
And every database developer should know everything about the relational algebra. Then there would no longer be a single developer left who had to post these stupid "I don't know how to do my job and want someone else to do it for me" questions on Stack Overflow anymore.
From my experience with relational databases, every developer should know:
- The different data types:
Using the correct type for the correct job will make your DB design more robust, your queries faster and your life easier.
- Learn about 1xM and MxM:
This is the bread and butter for relational databases. You need to understand one-to-many and many-to-many relations and apply then when appropriate.
- "K.I.S.S." principle applies to the DB as well:
Simplicity always works best. Provided you have studied how DB work, you will avoid unnecessary complexity which will lead to maintenance and speed problems.
- Indices:
It's not enough if you know what they are. You need to understand when to used them and when not to.
also:
Boolean algebra is your friend
Images: Don't store them on the DB. Don't ask why.
Test DELETE with SELECT
I would like everyone, both DBAs and developer/designer/architects, to better understand how to properly model a business domain, and how to map/translate that business domain model into both a normalized database logical model, an optimized physical model, and an appropriate object oriented class model, each one of which is (can be) different, for various reasons, and understand when, why, and how they are (or should be) different from one another.
I would say strong basic SQL skills. I've seen a lot of developers so far who know a little about databases but are always asking for tips about how to formulate a quite simple query. Queries are not always that easy and simple. You do have to use multiple joins (inner, left, etc.) when querying a well normalized database.
I think a lot of the technical details have been covered here and I don't want to add to them. The one thing I want to say is more social than technical, don't fall for the "DBA knowing the best" trap as an application developer.
If you are having performance issues with query take ownership of the problem too. Do your own research and push for the DBAs to explain what's happening and how their solutions are addressing the problem.
Come up with your own suggestions too after you have done the research. That is, I try to find a cooperative solution to the problem rather than leaving database issues to the DBAs.
Simple respect.
It's not just a repository
You probably don't know better than the vendor or the DBAs
You won't support it at 3 a.m. with senior managers shouting at you
Consider Denormalization as a possible angel, not the devil, and also consider NoSQL databases as an alternative to relational databases.
Also, I think the Entity-Relation model is a must-know for every developper even if you don't design databases. It'll let you understand thoroughly what's your database all about.
Never insert data with the wrong text encoding.
Once your database becomes polluted with multiple encodings, the best you can do is apply some kind combination of heuristics and manual labor.
Aside from syntax and conceptual options they employ (such as joins, triggers, and stored procedures), one thing that will be critical for every developer employing a database is this:
Know how your engine is going to perform the query you are writing with specificity.
The reason I think this is so important is simply production stability. You should know how your code performs so you're not stopping all execution in your thread while you wait for a long function to complete, so why would you not want to know how your query will affect the database, your program, and perhaps even the server?
This is actually something that has hit my R&D team more times than missing semicolons or the like. The presumtion is the query will execute quickly because it does on their development system with only a few thousand rows in the tables. Even if the production database is the same size, it is more than likely going to be used a lot more, and thus suffer from other constraints like multiple users accessing it at the same time, or something going wrong with another query elsewhere, thus delaying the result of this query.
Even simple things like how joins affect performance of a query are invaluable in production. There are many features of many database engines that make things easier conceptually, but may introduce gotchas in performance if not thought of clearly.
Know your database engine execution process and plan for it.
For a middle-of-the-road professional developer who uses databases a lot (writing/maintaining queries daily or almost daily), I think the expectation should be the same as any other field: You wrote one in college.
Every C++ geek wrote a string class in college. Every graphics geek wrote a raytracer in college. Every web geek wrote interactive websites (usually before we had "web frameworks") in college. Every hardware nerd (and even software nerds) built a CPU in college. Every physician dissected an entire cadaver in college, even if she's only going to take my blood pressure and tell me my cholesterol is too high today. Why would databases be any different?
Unfortunately, they do seem different, today, for some reason. People want .NET programmers to know how strings work in C, but the internals of your RDBMS shouldn't concern you too much.
It's virtually impossible to get the same level of understanding from just reading about them, or even working your way down from the top. But if you start at the bottom and understand each piece, then it's relatively easy to figure out the specifics for your database. Even things that lots of database geeks can't seem to grok, like when to use a non-relational database.
Maybe that's a bit strict, especially if you didn't study computer science in college. I'll tone it down some: You could write one today, completely, from scratch. I don't care if you know the specifics of how the PostgreSQL query optimizer works, but if you know enough to write one yourself, it probably won't be too different from what they did. And you know, it's really not that hard to write a basic one.
The order of columns in a non-unique index is important.
The first column should be the column that has the most variability in its content (i.e. cardinality).
This is to aid SQL Server ability to create useful statistics in how to use the index at runtime.
Understand the tools that you use to program the database!!!
I wasted so much time trying to understand why my code was mysteriously failing.
If you're using .NET, for example, you need to know how to properly use the objects in the System.Data.SqlClient namespace. You need to know how to manage your SqlConnection objects to make sure they are opened, closed, and when necessary, disposed properly.
You need to know that when you use a SqlDataReader, it is necessary to close it separately from your SqlConnection. You need to understand how to keep connections open when appropriate to how to minimize the number of hits to the database (because they are relatively expensive in terms of computing time).
Basic SQL skills.
Indexing.
Deal with different incarnations of DATE/ TIME/ TIMESTAMP.
JDBC driver documentation for the platform you are using.
Deal with binary data types (CLOB, BLOB, etc.)
For some projects, and Object-Oriented model is better.
For other projects, a Relational model is better.
The impedance mismatch problem, and know the common deficiencies or ORMs.
RDBMS Compatibility
Look if it is needed to run the application in more than one RDBMS. If yes, it might be necessary to:
avoid RDBMS SQL extensions
eliminate triggers and store procedures
follow strict SQL standards
convert field data types
change transaction isolation levels
Otherwise, these questions should be treated separately and different versions (or configurations) of the application would be developed.
Don't depend on the order of rows returned by an SQL query.
Three (things) is the magic number:
Your database needs version control too.
Cursors are slow and you probably don't need them.
Triggers are evil*
*almost always
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 9 years ago.
Improve this question
I've been hearing things about NoSQL and that it may eventually become the replacement for SQL DB storage methods due to the fact that DB interaction is often a bottle neck for speed on the web.
So I just have a few questions:
What exactly is it?
How does it work?
Why would it be better than using a SQL Database? And how much better is it?
Is the technology too new to start implementing yet or is it worth taking a look into?
There is no such thing as NoSQL!
NoSQL is a buzzword.
For decades, when people were talking about databases, they meant relational databases. And when people were talking about relational databases, they meant those you control with Edgar F. Codd's Structured Query Language. Storing data in some other way? Madness! Anything else is just flatfiles.
But in the past few years, people started to question this dogma. People wondered if tables with rows and columns are really the only way to represent data. People started thinking and coding, and came up with many new concepts how data could be organized. And they started to create new database systems designed for these new ways of working with data.
The philosophies of all these databases were different. But one thing all these databases had in common, was that the Structured Query Language was no longer a good fit for using them. So each database replaced SQL with their own query languages. And so the term NoSQL was born, as a label for all database technologies which defy the classic relational database model.
So what do NoSQL databases have in common?
Actually, not much.
You often hear phrases like:
NoSQL is scalable!
NoSQL is for BigData!
NoSQL violates ACID!
NoSQL is a glorified key/value store!
Is that true? Well, some of these statements might be true for some databases commonly called NoSQL, but every single one is also false for at least one other. Actually, the only thing NoSQL databases have in common, is that they are databases which do not use SQL. That's it. The only thing that defines them is what sets them apart from each other.
So what sets NoSQL databases apart?
So we made clear that all those databases commonly referred to as NoSQL are too different to evaluate them together. Each of them needs to be evaluated separately to decide if they are a good fit to solve a specific problem. But where do we begin? Thankfully, NoSQL databases can be grouped into certain categories, which are suitable for different use-cases:
Document-oriented
Examples: MongoDB, CouchDB
Strengths: Heterogenous data, working object-oriented, agile development
Their advantage is that they do not require a consistent data structure. They are useful when your requirements and thus your database layout changes constantly, or when you are dealing with datasets which belong together but still look very differently. When you have a lot of tables with two columns called "key" and "value", then these might be worth looking into.
Graph databases
Examples: Neo4j, GiraffeDB.
Strengths: Data Mining
While most NoSQL databases abandon the concept of managing data relations, these databases embrace it even more than those so-called relational databases.
Their focus is at defining data by its relation to other data. When you have a lot of tables with primary keys which are the primary keys of two other tables (and maybe some data describing the relation between them), then these might be something for you.
Key-Value Stores
Examples: Redis, Cassandra, MemcacheDB
Strengths: Fast lookup of values by known keys
They are very simplistic, but that makes them fast and easy to use. When you have no need for stored procedures, constraints, triggers and all those advanced database features and you just want fast storage and retrieval of your data, then those are for you.
Unfortunately they assume that you know exactly what you are looking for. You need the profile of User157641? No problem, will only take microseconds. But what when you want the names of all users who are aged between 16 and 24, have "waffles" as their favorite food and logged in in the last 24 hours? Tough luck. When you don't have a definite and unique key for a specific result, you can't get it out of your K-V store that easily.
Is SQL obsolete?
Some NoSQL proponents claim that their favorite NoSQL database is the new way of doing things, and SQL is a thing of the past.
Are they right?
No, of course they aren't. While there are problems SQL isn't suitable for, it still got its strengths. Lots of data models are simply best represented as a collection of tables which reference each other. Especially because most database programmers were trained for decades to think of data in a relational way, and trying to press this mindset onto a new technology which wasn't made for it rarely ends well.
NoSQL databases aren't a replacement for SQL - they are an alternative.
Most software ecosystems around the different NoSQL databases aren't as mature yet. While there are advances, you still haven't got supplemental tools which are as mature and powerful as those available for popular SQL databases.
Also, there is much more know-how for SQL around. Generations of computer scientists have spent decades of their careers into research focusing on relational databases, and it shows: The literature written about SQL databases and relational data modelling, both practical and theoretical, could fill multiple libraries full of books. How to build a relational database for your data is a topic so well-researched it's hard to find a corner case where there isn't a generally accepted by-the-book best practice.
Most NoSQL databases, on the other hand, are still in their infancy. We are still figuring out the best way to use them.
What exactly is it?
On one hand, a specific system, but it has also become a generic word for a variety of new data storage backends that do not follow the relational DB model.
How does it work?
Each of the systems labelled with the generic name works differently, but the basic idea is to offer better scalability and performance by using DB models that don't support all the functionality of a generic RDBMS, but still enough functionality to be useful. In a way it's like MySQL, which at one time lacked support for transactions but, exactly because of that, managed to outperform other DB systems. If you could write your app in a way that didn't require transactions, it was great.
Why would it be better than using a SQL Database? And how much better is it?
It would be better when your site needs to scale so massively that the best RDBMS running on the best hardware you can afford and optimized as much as possible simply can't keep up with the load. How much better it is depends on the specific use case (lots of update activity combined with lots of joins is very hard on "traditional" RDBMSs) - could well be a factor of 1000 in extreme cases.
Is the technology too new to start implementing yet or is it worth taking a look into?
Depends mainly on what you're trying to achieve. It's certainly mature enough to use. But few applications really need to scale that massively. For most, a traditional RDBMS is sufficient. However, with internet usage becoming more ubiquitous all the time, it's quite likely that applications that do will become more common (though probably not dominant).
Since someone said that my previous post was off-topic, I'll try to compensate :-) NoSQL is not, and never was, intended to be a replacement for more mainstream SQL databases, but a couple of words are in order to get things in the right perspective.
At the very heart of the NoSQL philosophy lies the consideration that, possibly for commercial and portability reasons, SQL engines tend to disregard the tremendous power of the UNIX operating system and its derivatives.
With a filesystem-based database, you can take immediate advantage of the ever-increasing capabilities and power of the underlying operating system, which have been steadily increasing for many years now in accordance with Moore's law. With this approach, many operating-system commands become automatically also "database operators" (think of "ls" "sort", "find" and the other countless UNIX shell utilities).
With this in mind, and a bit of creativity, you can indeed devise a filesystem-based database that is able to overcome the limitations of many common SQL engines, at least for specific usage patterns, which is the whole point behind NoSQL's philosophy, the way I see it.
I run hundreds of web sites and they all use NoSQL to a greater or lesser extent. In fact, they do not host huge amounts of data, but even if some of them did I could probably think of a creative use of NoSQL and the filesystem to overcome any bottlenecks. Something that would likely be more difficult with traditional SQL "jails". I urge you to google for "unix", "manis" and "shaffer" to understand what I mean.
If I recall correctly, it refers to types of databases that don't necessarily follow the relational form. Document databases come to mind, databases without a specific structure, and which don't use SQL as a specific query language.
It's generally better suited to web applications that rely on performance of the database, and don't need more advanced features of Relation Database Engines. For example, a Key->Value store providing a simple query by id interface might be 10-100x faster than the corresponding SQL server implementation, with a lower developer maintenance cost.
One example is this paper for an OLTP Tuple Store, which sacrificed transactions for single threaded processing (no concurrency problem because no concurrency allowed), and kept all data in memory; achieving 10-100x better performance as compared to a similar RDBMS driven system. Basically, it's moving away from the 'One Size Fits All' view of SQL and database systems.
In practice, NoSQL is a database system which supports fast access to large binary objects (docs, jpgs etc) using a key based access strategy. This is a departure from the traditional SQL access which is only good enough for alphanumeric values. Not only the internal storage and access strategy but also the syntax and limitations on the display format restricts the traditional SQL. BLOB implementations of traditional relational databases too suffer from these restrictions.
Behind the scene it is an indirect admission of the failure of the SQL model to support any form of OLTP or support for new dataformats. "Support" means not just store but full access capabilities - programmatic and querywise using the standard model.
Relational enthusiasts were quick to modify the defnition of NoSQL from Not-SQL to Not-Only-SQL to keep SQL still in the picture! This is not good especially when we see that most Java programs today resort to ORM mapping of the underlying relational model. A new concept must have a clearcut definition. Else it will end up like SOA.
The basis of the NoSQL systems lies in the random key - value pair. But this is not new. Traditional database systems like IMS and IDMS did support hashed ramdom keys (without making use of any index) and they still do. In fact IDMS already has a keyword NONSQL where they support SQL access to their older network database which they termed as NONSQL.
It's like Jacuzzi: both a brand and a generic name. It's not just a specific technology, but rather a specific type of technology, in this case referring to large-scale (often sparse) "databases" like Google's BigTable or CouchDB.
NoSQL the actual program appears to be a relational database implemented in awk using flat files on the backend. Though they profess, "NoSQL essentially has no arbitrary limits, and can work where other products can't. For example there is no limit on data field size, the number of columns, or file size" , I don't think it is the large scale database of the future.
As Joel says, massively scalable databases like BigTable or HBase, are much more interesting. GQL is the query language associated with BigTable and App Engine. It's largely SQL tweaked to avoid features Google considers bottle-necks (like joins). However, I haven't heard this referred to as "NoSQL" before.
NoSQL is a database system which doesn't use string based SQL queries to fetch data.
Instead you build queries using an API they will provide, for example Amazon DynamoDB is a good example of a NoSQL database.
NoSQL databases are better for large applications where scalability is important.
Does NoSQL mean non-relational database?
Yes, NoSQL is different from RDBMS and OLAP. It uses looser consistency models than traditional relational databases.
Consistency models are used in distributed systems like distributed shared memory systems or distributed data store.
How it works internally?
NoSQL database systems are often highly optimized for retrieval and appending operations and often offer little functionality beyond record storage (e.g. key-value stores). The reduced run-time flexibility compared to full SQL systems is compensated by marked gains in scalability and performance for certain data models.
It can work on Structured and Unstructured Data. It uses Collections instead of Tables
How do you query such "database"?
Watch SQL vs NoSQL: Battle of the Backends; it explains it all.