I'm trying to design a database schema for Djabgo rest framework web application.
At some point, I have two choces:
1- Choose a schema in which in one or several apies, I have to get a queryset from database and iterate and order it with python. (For example, I can store some datas in an array-data-typed column, get them from database and sort them with python.)
2- store the data in another table and insert a kind of big number of rows with each insert. This way, I can get the data in my favorite format in much less lines with orm codes.
I tried some basic tests and benchmarking to see which way is faster, and letting database handle more of the job (second way) didn't let me down. But I don't have the means of setting a more real situatuin and here's the question:
Is it still a good idea to let database handle the job when it also has to handle hundreds of requests from other apies and clients each second?
Is database (and orm) usually faster and more reliable than backend?
As a general rule, you want to let the database do work when the work is appropriate for the database. Sorting result sets would be in that category.
Keep in mind:
The database is running on a server, often on a distributed system and so it has access to more resources.
Databases are designed to handle large data, so they are not limited by the memory in a single thread.
When this question comes up, often more data needs to be passed back to the application than is strictly needed. Consider a problem such as getting the top 10 of something.
Mixing processing in the application and the database often requires multiple queries and passing data back and forth, which is expensive.
(And there are no doubt other considerations.)
There are some situations where it might be more efficient or convenient to do work in the application. A common example is formatting result sets for the application -- say turning 1234.56 into $1,234.56. Other examples would be when the application language has capabilities that are not directly in SQL or are hard to implement in SQL.
Related
I am working on a group project and we are having a discussion about whether to calculate data that we want from an existing database and store it in a new database to query later, or calculate the data from the existing database every time we need to use it. I was wondering what the pros and cons may be for either implementation. Is there any advice you could give?
Edit: Here is more elaborate explanation. We have a large database that has a lot of information being submitted to it daily. We are building a system to track certain points of data. For example, we are getting the count of how many times a user does something that is entered in the database. Using this example (are actual idea is a bit more complex), we are discussing to methods of getting the count of actions per users. The first method is to create a database that stores the users and their action count, and query this database every time we need the action count. The second method would be to query the large database and count the actions per user every time we need to use it. I hope this explanation helps explain. Thoughts?
Edit 2: Two more things that may be useful to point out is 1: I only have read access to the large database and 2: My ultimate goal is to display this information on a web page for end users.
This is a generic question about optimization by caching. The following was my answer to essentially the same question. Even though that question provided a bunch of different details, none of them were specific enough to merit a non-generic answer either:
The more you want to calculate at query time, the more you want views,
calculated columns and stored or user routines. The more you want to
calculate at normalized base update time, the more you want cascades
and triggers. The more you want to calculate at some other (scheduled
or ad hoc) time, the more you use snapshots aka materialized views and
updated denormalized bases. You can combine these. Any time the
database is accessed it can be enabled by and restricted by stored
routines or other api.
Until you can show that they are in adequate, views and calculated
columns are the simplest.
The whole idea of a DBMS is to store a representation of your
application state as the database (which normalization reduces the
redundancy of) and then you query and let the DBMS implement and
optimize calculation of the answer. You haven't presented a reason for
not doing that in the most straightforward way possible.
[sic]
Always make sure an application is reading its own personal ("external") database that is a view of "the" ("conceptual") database so that when you change the implemention of the former (plus the rest of some combined interfact) by the latter (plus the rest of some compbined mechanisms) your applications do not have to change ("logical independence"). Here the applications are your users' and your trackers'.
Ultimately you must instrument and guestimate. When it is worth it you start caching. Preferably as much as possible in terms of high-level notions like views and snapshots and as little as possible in non-DBMS code. One of he benefits of the relational model is that it is easy to describe a strightforward relational interface in terms of another straightforward relational interface. You protect your applications from change by offering an interface that hides secrets of implementation or which of a family of interfaces is the current one.
I'm working for a company running a software product based on a MS SQL database server, and through the years I have developed 20-30 quite advanced reports in PHP, taking data directly from the database. This has been very successful, and people are happy with it.
But it has some drawbacks:
For new changes, it can be quite development intensive
The user can't experiment much with the data - it is locked to a hard-coded view
It can be slow for big reports
I am considering gradually going to a OLAP-based approach, which can be queried from Excel or some web-based service. But I would like to do this in a way that introduces the least amount of new complexity in the IT environment - the least amount of different services, synchronization jobs etc!
I have some questions in this regard:
1) Workflow-related:
What is a good development route from "black box SQL server" to "OLAP ready to use"?
Which servers and services should be set up, and which scripts should be written?
Which are the hardest/most critical/most time-intensive parts?
2) ETL:
I suppose it is best to have separate servers for their Data Warehouse and Production SQL?
How are these kept in sync (push/pull)? Using which technologies/languages?
For me SSIS looks overly complicated, and the graphical workflow doesn't appeal much to me -- I would rather like a text based script that does the job. Is this feasible?
Or is it advantagous to use the graphical client with only one source and one destination?
3) Development:
How much of this (data integration, analysis services) can be efficiently maintained from a CLI-tool?
Can the setup be transferred back and forth between production and development easily?
I'm happy with any answer that covers just some of this - and even though it is a MS environment, I'm also interested to hear about advantages in other technologies.
I only have experience with Microsoft OLAP, so here are my two cents regarding what I know:
If you are implementing cubes, then separate the production SQL Server from the source for the cubes. Cubes require a lot of SELECT DISTINCT column_name FROM source.table. You don't want cube processing to block your mission critical production system.
Although you can implement OLAP cubes with standard relation tables, you will quickly find that unless your data is a ledger-style system you will probably need to fully reprocess your fact and dimension tables and this will require requerying the source database over and over again. That's a large argument for building a separate data warehouse that uses ledger-style transactions for the fact tables. For instance, if a customer orders something and then cancels it, your source system may track this as a status change. In your fact table, you probably need to show this as a row for ordering that has a positive quantity and revenue stream and a row for cancelling that has a negative quantity and revenue stream.
OLAP may be overkill for your environment. The main issue you appeared to raise was that your reports are static and users want access to the data directly. You could build a data model and give users Report Builder access in SSRS, or report writing access in some other BI suite like Cognos, Business Objects, etc. I don't generally recommend this approach since it is way beyond what most users should have to know to get data, but in a small shop this may be sufficient and it is easy to implement. Let's face it -- users generally just want to get the data into Excel to manipulate it further. So if you don't want to give them a web front-end and you just want them to get to the data from Excel, you could give them direct database access to a copy of the production data. The downside of this approach is users don't generally understand SQL or database relationships. OLAP helps you avoid forcing users to learn SQL or relationships, but is isn't easy to implement on your end. If you only have a couple of power users who need this kind of access, it could be easy enough to teach the few power users how to do basic queries in Excel against the database and they will be happy to get this tomorrow. OLAP won't be ready by tomorrow.
If you only have a few kinds of source data systems, you could get away with building a super-dynamic static report. For instance, I have a report that was written in C# that basically allows users to select as many columns as they want from a list of 30 columns and filter the data on a few date range fields and field filter lists. This simple report covers about 40% of all ad hoc report requests from end-users since it covers all the basic, core customer metrics and fields. We recently moved this report to SSRS and that allowed us to up the number of fields to about 100 and improved the overall user experience. Regardless of the reporting platform, it is possible to give users some dynamic flexibility even in the confines of a static reporting system.
If you only have a couple of databases, you can probably backup and restore the databases as your ETL. However, if you want to do anything beyond that, then you might as well bite the bullet and use SSIS (or some other ETL tool). Once you get into ETL for data warehousing, you are going to use a graphic-oriented design tool. Coding works well for applications, but ETL is more about workflows and that's why the tools tend to converge on a graphical UI. You can work around this and try to code a data warehouse from a text editor, but in the end you are going to lose out on a lot. See this post for more details on the differences between loading data from code and loading data from SSIS.
FEEDBACK ON HOW TO USE CUBES WITH A RELATIONAL DATA STORE
It is possible to implement a cube over a relational data store, but there are some major problems with using this approach. The main reason it is technically feasible has to do with how you configure your DSV. The DSV is essentially a logical layer between the physical database and the cube/dimension definitions. Instead of importing the relational tables into the DSV, you could define Named Queries or create views in the database that flatten the data.
The advantage of this approach are as follows:
It is relatively easy to implement since you don't have to build an entire ETL subsystem to get started with OLAP.
This approach works well for prototyping how you want to build a more long-term solution. You can prototype it in 1-2 days and show some of the benefits of OLAP today.
Some very, very large tables don't have to be completely duplicated just to support an OLAP cube. I have several multi-billion row tables that are almost completely standardized fact tables. The only columns they don't have are date keys and they also contain some NULL values on fields that shouldn't have nulls at all. Instead of duplicating these very massive tables, you can create the surrogate date keys and set values for the nulls in the view or named query. If you aren't going to see a huge performance boon for duplicating the table, then this may be a candidate for leaving in a more raw format in the database itself.
The disadvantages of this approach are as follows:
If you haven't built a true Kimball method data warehouse, then you probably aren't tracking transactions in a ledger-style. Kimball method fact tables (at least as I understand them) always change values by adding and subtracting rows. If someone cancels part of an order, you can't update the value in the cube for the single transaction. Instead, you have to balance out the transaction with a negative value. If you have to update the transaction, then you will have to fully reprocess the partition of the cube to replace the value which can be a very expensive operation. Unless your source system is a ledger-style transaction system, you will probably have to build a ledger-style copy in your ETL subsystem.
If you don't build a Kimball method data warehouse, then you are probably using unobscured and possibly non-integer primary keys in your database. This directly impacts query performance inside the cube. It also sets you up for having a theoretically inflexible data warehouse. For instance, if you have an product ordering system that uses an integer key and you start using a second product ordering system either as a replacement for the legacy system or in tandem with the legacy system, you may struggle to combine the data together merely through the DSV since each system has different data points, metrics, workflows, data types, etc. Worse, if they have the same data types for the order id and the order id values overlap between systems, then you must declare a surrogate key that you can use across both systems. This can be difficult, but not impossible, to implement without using a flattened data warehouse.
You may have to build the system twice if you start with the relational data store and then move to flattened database. Frankly, I think the amount of duplicated work is trivial. Most of what you learned building the cube off a relational data store will translate to setting up the new OLAP cube. The main problem, though, is that you will probably create a new cube altogether and then any users of the old cube will have to migrate to the new cube. Any reports built in SSRS or Excel will probably break at that point and need to be rewritten from the ground up. So the main cost of rebuilding the cube is really on rebuilding dependent objects -- not on rebuilding the cube itself.
Let me know if you want me to expand on any of the above points. good luck.
You're basically asking the million dollar question of "How do I build a DWH". This is not really a question that can decisively be answered.
Nevertheless, here is a kickstart:
If you are looking for a minimum viable product, be aware that you are in a data environment, and not a pure software one. In data-heavy environments, it is much harder to incrementally build a product, because the amount of effort to introduce changes in the system is much greater. Think about it as if every change you make in a piece of software has to be somehow backwards-compatible with anything you've ever done. Now you understand the hell Microsoft are in :-).
Also, data systems involve many third-party tools such as DBs, ETL tools and reporting platforms. The choices you make should be viable for the expected development of your system, else you might have to completely replace these tools down the road.
While you can start with a DB cloning that will be based on simple copy SQLs and then aggregating it or pushing it into an OLAP, I would recommend getting your hands dirty with a real ETL tool from the start. This is especially true if you foresee the need to grow. 9 out of 10 times, the need will grow.
MS-SQL is a good choice for a DB if you don't mind the cost. The natural ETL tool would be SSIS, and it's a solid tool as well.
Even if your first transformations are merely "take this table and dump it in there", you still gain a lot in terms of process management (has the job run? What happens if it fails? etc) and debugging. Also, it is easier to organically grow as requirements and/or special cases have to be dealt with.
In our web application we need to trace what users click, what they write into search box, etc. Lots of data will be sent by AJAX. Generally functionality is a bit similar to google analytics, but we need to customize it in different ways.
Data will be collected and once per day aggregated and exported to PostgreSQL, so backend should be able to handle dozens of inserts. I don't consider usage of traditional SQL database, because probably it won't handle so many inserts efficiently.
I wonder which backend would you use for such task? Actually I think about MongoDB or Cassandra. But maybe you know better software for that task? Maybe something different then NoSQL database?
Web application is written in Ruby on Rails so support for Ruby would be nice but that's definitely not the most important.
Sounds like you need to analyse your specific requirements.
It may be that the best solution is to split / partition / shard a conventional database and then push the data up from there.
Depending on what your tolerance for data loss is, there are a lot of options. If you choose a system which has single-server durability, a major source of write bottleneck will be fdatasync() (assuming you use hard drives to store your data on).
If you can tolerate syncing less often than on every commit, then you may be able to tune your database to commit at timed intervals.
Depending on your table, index structure etc, I'd expect that you can get rather a lot of inserts with a "conventional" db (e.g. postgresql), if you manage it correctly and tune the durability (if it supports that) to your liking.
Sharding this into several instances of course will enable you to scale this up. However, you need to be mindful of operational requirements (i.e. what happens if some of the instances are down). Talk to your Ops team about what they're comfortable managing.
For typical 3-tiered application, I have seen that in many cases they use a lot of complex stored procedures in the database. I cannot quite get the benefit of this approach. In my personal understanding, there are following disadvantages on this approach:
Transactions become coarse.
Business logic goes into database.
Lots of computation is done in the database server, rather than in the application server. Meanwhile, the database still needs to do its original work: maintain data. The database server may become a bottleneck.
I can guess there may be 2 benefits of it:
Change the business logic without compile. But the SPs are much more harder to maintain and test than Java/C# code.
Reduce the number of DB connection. However, in the common case, the bottleneck of database is hard disk io rather than network io.
Could anyone please tell me the benefits of using a lot of stored procedures rather than letting the work be done in business logic layer?
Basically, the benefit is #2 of your problem list - if you do a lot of processing in your database backend, then it's handled there and doesn't depend on the application accessing the database.
Sure - if your application does all the right things in its business logic layer, things will be fine. But as soon as a second and a third application need to connect to your database, suddenly they too have to make sure to respect all the business rules etc. - or they might not.
Putting your business rules and business logic in the database ensures that no matter how an app, a script, a manager with Excel accesses your database, your business rules will be enforced and your data integrity will be protected.
That's the main reason to have stored procs instead of code-based BLL.
Also, using Views for read and Stored Procs for update/insert, the DBA can remove any direct permissions on the underlying tables. Your users do no longer need to have all the rights on the tables, and thus, your data in your tables is better protected from unadvertent or malicious changes.
Using a stored proc approach also gives you the ability to monitor and audit database access through the stored procs - no one will be able to claim they didn't alter that data - you can easily prove it.
So all in all: the more business critical your data, the more protection layer you want to build around it. That's what using stored procs is for - and they don't need to be complex, either - and most of those stored procs can be generated based on table structure using code generation, so it's not a big typing effort, either.
Don't fear the DB.
Let's also not confuse business logic with data logic which has its rightful place at the DB.
Good systems designers will encompass flexible business logic through data logic, i.e. abstract business rule definitions which can be driven by the (non)existence or in attributes of data rows.
Just FYI, the most successful and scalable "enterprise/commercial" software implementations with which I have worked put all projection queries into views and all data management either into DB procedures or triggers on staged tables.
Network between appServer and sqlServer is the bottle neck very often.
Stored procedures are needed when you need to do complex query.
For example you want collect some data about employee by his surname. Especially imagine, that data in DB looks like some kind of tree - you have 3 records about this employee in table A. You have 10 records in table B for each record in table A. You have 100 records in table C for each record in table B. And you want to get only special 5 records from table C about that employee. Without stored procedures you will get a lot of queries traffic between appServer and sqlServer, and a lot of code in appServer. With stored procedure which accepts employee surname, fetches those 5 records and returns them to appServer you 1) decrease traffic by hundreds times, 2) greatly simplify appServer code.
The life time of our data exceeds that of our applications. Also data gets shared between applications. So many applications will insert data into the database, many applications will retrieve data from it. The database is responsible for the completeness, integrity and correctness of the data. Therefore it needs to have the authority to enforce the business rules relating to the data.
Taking you specific points:
Transactions are Units Of Work. I
fail to see why implementing
transactions in stored procedures
should change their granularity.
Business logic which applies to the
data belongs with the data: that
maximises cohesion.
It is hard to write good SQL and to
learn to think in sets. Therefore
it may appear that the database is
the bottleneck. In fact, if we are
undertaking lots of work which
relates to the data the database is
probably the most efficient place to
do.
As for maintenance: if we are familiar with PL/SQL, T-SQL, etc maintenance is easier than it might appear from the outside. But I concede that tool support for things like refactoring lags behind that of other languages.
You listed one of the main ones putting business logic in the Db often gives the impression of making it easier to maintain.
Generally complex SP logic in the db allows for cheaper implementation of the actual implementation code, which may be beneficial if its a transitional application (say being ported from legacy code), its code which needs to be implemented in several languages (for instance to market on different platforms or devices) or because the problem is simpler to solve in the db.
One other reason for this is often there is a general "best practice" to encapsulate all access to the db in sps for security or performance reasons. Depending on your platform and what you are doing with it this may or may not be marginally true.
I don't think there are any. You are very correct that moving the BL to the database is bad, but not for everything. Try taking a look at Domain Driven Design. This is the antidote to massive numbers of SPROCs. I think you should be using your database as somewhere to store you business objects, nothing more.
However, SPROCs can be much more efficient on certain, simple functions. For instance, you might want to increase the salary to every employee in your database by a fixed percentage. This is quicker to do via a SPROC than getting all the employees from the db, updating them and then saving them back.
I worked in a project where every thing is literally done in database level. We wrote lot of stored procedures and did lot of business validation / logic in the database. Most of the times it became a big overhead for us to debug.
The advantages I felt may be
Take advantage of full DB features.
Database intense activities like lot of insertion/updation can be better done in DB level. Call an SP and let it do all the work instead of hitting DB several times.
New DB servers can accommodate complex operations so they no longer see this as a bottleneck. Oh yeah, we used Oracle.
Looking at it now, I think few things could have been better done at application level and lesser at DB level.
It depends almost entirely on the context.
Doing work on the server rather than on the clients is generally a bad idea as it makes your server less scalable. However, you have to balance this against the expected workload (if you know you will only have 100 users in a closed enironment, you may not need a scalable server) and network traffic costs (if you have to read a lot of data to apply calculations/processes to, then it can be cheaper/faster overall to run those calculations on the server and only send the results over the net).
Also, if you have custom client applications (as opposed to web browsers etc) it makes it very easy to push updates out to your clients, because you don't need to recompile and deploy the client code, you simply upgrade the database stored procedures.
Of course, using stored procedures rather than executing dynamically compiled SQL statements can be more efficient (it's precompiled, and the code doesn't need to be uploaded to the server) and aids encapsulation to give the database better integrity/security. But by the sound of it, you're talking about masses of busines logic, not simple efficiency and security measures.
As with most things, a sensible compromise/balance is needed. Stored Procedures should be used enough to enhance efficiency and security, but you don't want your server to become unscalable.
"there are following disadvantages on this approach:
...
Business logic goes into database."
Insofar as by "busines logic" you mean "enforcement of business rules", the DBMS is EXACTLY where "business logic" belongs.
I'm working on a web app that is somewhere between an email service and a social network. I feel it has the potential to grow really big in the future, so I'm concerned about scalability.
Instead of using one centralized MySQL/InnoDB database and then partitioning it when that time comes, I've decided to create a separate SQLite database for each active user: one active user per 'shard'.
That way backing up the database would be as easy as copying each user's small database file to a remote location once a day.
Scaling up will be as easy as adding extra hard disks to store the new files.
When the app grows beyond a single server I can link the servers together at the filesystem level using GlusterFS and run the app unchanged, or rig up a simple SQLite proxy system that will allow each server to manipulate sqlite files in adjacent servers.
Concurrency issues will be minimal because each HTTP request will only touch one or two database files at a time, out of thousands, and SQLite only blocks on reads anyway.
I'm betting that this approach will allow my app to scale gracefully and support lots of cool and unique features. Am I betting wrong? Am I missing anything?
UPDATE I decided to go with a less extreme solution, which is working fine so far. I'm using a fixed number of shards - 256 sqlite databases, to be precise. Each user is assigned and bound to a random shard by a simple hash function.
Most features of my app require access to just one or two shards per request, but there is one in particular that requires the execution of a simple query on 10 to 100 different shards out of 256, depending on the user. Tests indicate it would take about 0.02 seconds, or less, if all the data is cached in RAM. I think I can live with that!
UPDATE 2.0 I ported the app to MySQL/InnoDB and was able to get about the same performance for regular requests, but for that one request that requires shard walking, innodb is 4-5 times faster. For this reason, and other reason, I'm dropping this architecture, but I hope someone somewhere finds a use for it...thanks.
The place where this will fail is if you have to do what's called "shard walking" - which is finding out all the data across a bunch of different users. That particular kind of "query" will have to be done programmatically, asking each of the SQLite databases in turn - and will very likely be the slowest aspect of your site. It's a common issue in any system where data has been "sharded" into separate databases.
If all the of the data is self-contained to the user, then this should scale pretty well - the key to making this an effective design is to know how the data is likely going to be used and if data from one person will be interacting with data from another (in your context).
You may also need to watch out for file system resources - SQLite is great, awesome, fast, etc - but you do get some caching and writing benefits when using a "standard database" (i.e. MySQL, PostgreSQL, etc) because of how they're designed. In your proposed design, you'll be missing out on some of that.
Sounds to me like a maintenance nightmare. What happens when the schema changes on all those DBs?
http://freshmeat.net/projects/sphivedb
SPHiveDB is a server for sqlite database. It use JSON-RPC over HTTP to expose a network interface to use SQLite database. It supports combining multiple SQLite databases into one file. It also supports the use of multiple files. It is designed for the extreme sharding schema -- one SQLite database per user.
One possible problem is that having one database for each user will use disk space and RAM very inefficiently, and as the user base grows the benefit of using a light and fast database engine will be lost completely.
A possible solution to this problem is to create "minishards" consisting of maybe 1024 SQLite databases housing up to 100 users each. This will be more efficient than the DB per user approach, because data is packed more efficiently. And lighter than the Innodb database server approach, because we're using Sqlite.
Concurrency will also be pretty good, but queries will be less elegant (shard_id yuckiness). What do you think?
If you're creating a separate database for each user, it sounds like you're not setting up relationships... so why use a relational database at all?
If your data is this easy to shard, why not just use a standard database engine, and if you scale large enough that the DB becomes the bottleneck, shard the database, with different users in different instances? The effect is the same, but you're not using scores of tiny little databases.
In reality, you probably have at least some shared data that doesn't belong to any single user, and you probably frequently need to access data for more than one user. This will cause problems with either system, though.
I am considering this same architecture as I basically wanted to use the server side SQLLIte databases as backup and synching copy for clients. My idea for querying across all the data is to use Sphinx for full-text search and run Hadoop jobs from flat dumps of all the data to Scribe and then expose the results as webservies. This post gives me some pause for thought however, so I hope people will continue to respond with their opinion.
Having one database per user would make it really easy to restore individual users data of course, but as #John said, schema changes would require some work.
Not enough to make it hard, but enough to make it non-trivial.