What is the best approach for Sharding MySQL tables.
The approaches I can think of are :
Application Level sharding?
Sharding at MySQL proxy layer?
Central lookup server for sharding?
Do you know of any interesting projects or tools in this area?
The best approach for sharding MySQL tables to not do it unless it is totally unavoidable to do it.
When you are writing an application, you usually want to do so in a way that maximizes velocity, developer speed. You optimize for latency (time until the answer is ready) or throughput (number of answers per time unit) only when necessary.
You partition and then assign partitions to different hosts (= shard) only when the sum of all these partitions does no longer fit onto a single database server instance - the reason for that being either writes or reads.
The write case is either a) the frequency of writes is overloading this servers disks permanently or b) there are too many writes going on so that replication permanently lags in this replication hierarchy.
The read case for sharding is when the size of the data is so large that the working set of it no longer fits into memory and data reads start hitting the disk instead of being served from memory most of the time.
Only when you have to shard you do it.
The moment you shard, you are paying for that in multiple ways:
Much of your SQL is no longer declarative.
Normally, in SQL you are telling the database what data you want and leave it to the optimizer to turn that specification into a data access program. That is a good thing, because it is flexible, and because writing these data access programs is boring work that harms velocity.
With a sharded environment you are probably joining a table on node A against data on node B, or you have a table larger than a node, on nodes A and B and are joining data from it against data that is on node B and C. You are starting to write application side hash-based join resolutions manually in order to resolve that (or you are reinventing MySQL cluster), meaning you end up with a lot of SQL that no longer declarative, but is expressing SQL functionality in a procedural way (e.g. you are using SELECT statements in loops).
You are incurring a lot of network latency.
Normally, an SQL query can be resolved locally and the optimizer knows about the costs associated with local disk accesses and resolves the query in a way that minimizes the costs for that.
In a sharded environment, queries are resolved by either running key-value accesses across a network to multiple nodes (hopefully with batched key accesses and not individual key lookups per round trip) or by pushing parts of the WHERE clause onward to the nodes where they can be applied (that is called 'condition pushdown'), or both.
But even in the best of cases this involves many more network round trips that a local situation, and it is more complicated. Especially since the MySQL optimizer knows nothing about network latency at all (Ok, MySQL cluster is slowly getting better at that, but for vanilla MySQL outside of cluster that is still true).
You are losing a lot of expressive power of SQL.
Ok, that is probably less important, but foreign key constraints and other SQL mechanisms for data integrity are incapable of spanning multiple shards.
MySQL has no API which allows asynchronous queries that is in working order.
When data of the same type resides on multiple nodes (e.g. user data on nodes A, B and C), horizontal queries often need to be resolved against all of these nodes ("Find all user accounts that have not been logged in for 90 days or more"). Data access time grows linearly with the number of nodes, unless multiple nodes can be asked in parallel and the results aggregated as they come in ("Map-Reduce").
The precondition for that is an asynchronous communication API, which does not exist for MySQL in a good working shape. The alternative is a lot of forking and connections in the child processes, which is visiting the world of suck on a season pass.
Once you start sharding, data structure and network topology become visible as performance points to your application. In order to perform reasonably well, your application needs to be aware of these things, and that means that really only application level sharding makes sense.
The question is more if you want to auto-shard (determining which row goes into which node by hashing primary keys for example) or if you want to split functionally in a manual way ("The tables related to the xyz user story go to this master, while abc and def related tables go to that master").
Functional sharding has the advantage that, if done right, it is invisible to most developers most of the time, because all tables related to their user story will be available locally. That allows them to still benefit from declarative SQL as long as possible, and will also incur less network latency because the number of cross-network transfers is kept minimal.
Functional sharding has the disadvantage that it does not allow for any single table to be larger than one instance, and it requires manual attention of a designer.
Functional sharding has the advantage that it is relatively easily done to an existing codebase with a number of changes that is not overly large. http://Booking.com has done it multiple times in the past years and it worked well for them.
Having said all that, looking at your question, I do believe that you are asking the wrong questions, or I am completely misunderstanding your problem statement.
Application Level sharding: dbShards is the only product that I know of that does "application aware sharding". There are a few good articles on the website. Just by definition, application aware sharding is going to be more efficient. If an application knows exactly where to go with a transaction without having to look it up or get redirected by a proxy, that in its self will be faster. And speed is often one of the primary concerns, if not the only concern, when someone is looking into sharding.
Some people "shard" with a proxy, but in my eyes that defeats the purpose of sharding. You are just using another server to tell your transactions where to find the data or where to store it. With application aware sharding, your application knows where to go on its own. Much more efficient.
This is the same as #2 really.
Do you know of any interesting projects or tools in this area?
Several new projects in this space:
citusdata.com
spockproxy.sourceforge.net
github.com/twitter/gizzard/
Application level of course.
Best approach I've ever red I've found in this book
High Performance MySQL
http://www.amazon.com/High-Performance-MySQL-Jeremy-Zawodny/dp/0596003064
Short description: you could split your data in many parts and store ~50 part on each server. It will help you to avoid the second biggest problem of sharding - rebalancing. Just move some of them to the new server and everything will be fine :)
I strongly recommend you to buy it and read "mysql scaling" part.
As of 2018, there seems to be a MySql-native solution to that. There are actually at least 2 - InnoDB Cluster and NDB Cluster(there is a commercial and a community version of it).
Since most people who use MySql community edition are more familiar with InnoDB engine, this is what should be explored as a first priority. It supports replication and partitioning/sharding out of the box and is based on MySql Router for different routing/load-balancing options.
The syntax for your tables creation would need to change, for example:
CREATE TABLE t1 (col1 INT, col2 CHAR(5), col3 DATETIME) PARTITION BY HASH ( YEAR(col3) );
(this is only one of four partitioning types)
One very important limitation:
InnoDB foreign keys and MySQL partitioning are not compatible. Partitioned InnoDB tables cannot have foreign key references, nor can they have columns referenced by foreign keys. InnoDB tables which have or which are referenced by foreign keys cannot be partitioned.
Shard-Query is an OLAP based sharding solution for MySQL. It allows you to define a combination of sharded tables and unsharded tables. The unsharded tables (like lookup tables) are freely joinable to sharded tables, and sharded tables may be joined to each other as long as the tables are joined by the shard key (no cross shard or self joins that cross shard boundaries). Being an OLAP solution, Shard-Query usually has minimum response times of 100ms or less, even for simple queries so it will not work for OLTP. Shard-Query is designed for analyzing big data sets in parallel.
OLTP sharding solutions exist for MySQL as well. Closed source solutions include ScaleDB, DBShards. Open source OLTP solution include JetPants, Cubrid or Flock/Gizzard (Twitter infrastructure).
Do you know of any interesting projects or tools in this area?
As of 2022 Here are 2 tools:
Vitess (website: https://vitess.io & repo: https://github.com/vitessio/vitess)
PlanetScale (https://planetscale.com)
You can consider this middleware
shardingsphere
Related
In MySQL (and PostgreSQL), what exactly constitutes a DB instance and a DB partition?
For example, do different DB partitions need to necessarily live on different database instances? Or can a single DB instance manage multiple partitions? If the latter, what's the point of calling it a "partition"? Would the DB have any knowledge of it in this case?
Here's a quote from a document describing a system design from an online course:
How can we plan for the future growth of our system?
We can have a large number of logical partitions to accommodate future data growth, such that in the beginning, multiple logical partitions reside on a single physical database server. Since each database server can have multiple database instances on it, we can have separate databases for each logical partition on any server. So whenever we feel that a particular database server has a lot of data, we can migrate some logical partitions from it to another server. We can maintain a config file (or a separate database) that can map our logical partitions to database servers; this will enable us to move partitions around easily. Whenever we want to move a partition, we only have to update the config file to announce the change.
These terms are confusing, misused, and inconsistently defined.
For MySQL:
A Database has multiple definitions:
A "schema" (as used by other vendors/standards). This is a collection of tables. There are one or more "databases in an instance.
The instance. You should use "server" or "database server" to be clearer.
The data. "Dataset" might be a better term.
An instance refers to a copy of mysqld running on some machine somewhere.
You can have multiple instances on a single piece of hardware. (Rare)
You can have multiple instances on a single piece of hardware, with the instances in different VMs or Dockers. (handy for testing)
Usually "instance" refers to one server with one copy of MySQL on it. (Typical for larger-scale situations)
A PARTITION is a specific way to lay out a table (in a database).
It is seen in CREATE TABLE (...) PARTITION BY ....
It is a "horizontal" split of the data, often by date, but could be by some other 'column'.
It have no direct impact on performance, making it rarely useful.
Sharding is not implemented in MySQL, but can be done on top of MySQL.
It is also a "horizontal" split of the data, but in this case across multiple "instances".
The use case is, for example, social media where there are millions of "users" that are mostly handled by themselves. That is, most of the queries focus on a single slice of the data, hence it is practical to a bunch of users on one server and do all those queries there.
It can be called "horizontal partitioning" but should not be confused with PARTITIONs of a table.
Vertical partitioning is where some columns are pulled out of a table in put into a parallel table.
Both tables would (normally) have the same PRIMARY KEY, thereby facilitating JOINs.
Vertical partitioning would (normally) be done only in a single "instance".
The purposes include splitting off big text/blog columns; splitting off optional columns (and use LEFT JOIN to get NULLs).
Vertical partitioning was somewhat useful in MyISAM, but rarely useful in InnoDB, since that engine automatically does such.
Replication and Clustering
Multiple instances contain the same data.
Used for "High Availability" (HA).
Used for scaling out reads.
Orthogonally to partitioning or sharding.
Does not make sense to have the instances on the same server (except for testing/experimenting/staging/etc).
Partitions, in terms of MySQL and PostgreSQL feature set, are physical segmentations of data. They exist within a single database instance, and are used to reduce the scope of data you're interacting with at a particular time, to cope with high data volume situations.
The document you're quoting from is speaking of a more abstract concept of a data partition at the system design level.
MySQL temporary table are stored in memory as long as computer has enough RAM (and MySQL was set up accordingly). One can created any indexes for any fields.
Redis stores data in memory indexed by one key at time and in my understanding MySQL can do this job too.
Are there any things that make Redis better for storing big amount(100-200k rows) of volatile data? I can only explain the appearance of Redis that not every project has mysql inside and probably some other databases don't support temporary tables.
If I already have MySql in my project, does it make sense to put up with Redis?
Redis is like working with indexes directly. There's no ACID, SQL parser and many other things between you and the data.
It provides some basic data structures and they're specifically optimized to be held in memory, and they also have specific operations to read and modify them.
In the other hand, Redis isn't designed to query data (but you can implement very powerful and high-performant filters with SORT, SCAN, intersections and other operations) but to store the data as you're going to be consumed later. If you want to get, for example, customers sorted by 3 different criterias, you'll need to work to fill 3 different sorted sets. There're a lot of use cases with other data structures, but I would end up writing a book in an answer...
Also, one of most powerful features found in Redis is how easy can be replicated, and since its 3.0 version, it supports data sharding out-of-the-box.
About why you would need to use Redis instead of temporary tables on MySQL (and other engines which have them too) is up to you. You need to study your case and check if caching or storing data in a NoSQL storage like Redis can both outperform your actual approach and it provides you a more elegant data architecture.
By using Redis alongside the other database, you're effectively reducing the load on it. Also, when Redis is running on a different server, scaling can be performed independently on each tier.
I am looking for a free SQL database able to handle my data model. The project is a production database working in a local network not connected to the internet without any replication. The number of application connected at the same times would be less than 10.
The data volume forecast for the next 5 years are:
3 tables of 100 millions rows
2 tables of 500 millions rows
20 tables with less than 10k rows
My first idea was to use MySQL, but I have found around the web several articles saying that MySQL is not designed for big database. But, what is the meaning of big in this case?
Is there someone to tell me if MySQL is able to handle my data model?
I read that Postgres would be a good alternative, but require a lot of hours for tuning to be efficient with big tables.
I don't think so that my project would use NOSQL database.
I would know if someone has some experience to share with regarding MySQL.
UPDATE
The database will be accessed by C# software (max 10 at the same times) and web application (2-3 at the same times),
It is important to mention that only few update will be done on the big tables, only insert query. Delete statements will be only done few times on the 20 small tables.
The big tables are very often used for select statement, but the most often in the way to know if an entry exists, not to return grouped and ordered batch of data.
I work for Percona, a company that provides consulting and other services for MySQL solutions.
For what it's worth, we have worked with many customers who are successful using MySQL with very large databases. Terrabytes of data, tens of thousands of tables, tables with billions of rows, transaction load of tens of thousands of requests per second. You may get some more insight by reading some of our customer case studies.
You describe the number of tables and the number of rows, but nothing about how you will query these tables. Certainly one could query a table of only a few hundred rows in a way that would not scale well. But this can be said of any database, not just MySQL.
Likewise, one could query a table that is terrabytes in size in an efficient way. It all depends on how you need to query it.
You also have to set specific goals for performance. If you want queries to run in milliseconds, that's challenging but doable with high-end hardware. If it's adequate for your queries to run in a couple of seconds, you can be a lot more relaxed about the scalability.
The point is that MySQL is not a constraining factor in these cases, any more than any other choice of database is a constraining factor.
Re your comments.
MySQL has referential integrity checks in its default storage engine, InnoDB. The claim that "MySQL has no integrity checks" is a myth often repeated over the years.
I think you need to stop reading superficial or outdated articles about MySQL, and read some more complete and current documentation.
MySQLPerformanceBlog.com
High Performance MySQL, 3rd edition
MySQL 5.6 manual
MySQL has a two important (and significantly different) database engines - MyISAM and InnoDB. A limits depends on usage - MyISAM is nontransactional - there is relative fast import, but it is too simple (without own memory cache) and JOINs on tables higher than 100MB can be slow (due too simple MySQL planner - hash joins is supported from 5.6). InnoDB is transactional and is very fast on operations based on primary key - but import is slower.
Current versions of MySQL has not good planner as Postgres has (there is progress) - so complex queries are usually much better on PostgreSQL - and really simple queries are better on MySQL.
Complexity of PostgreSQL configuration is myth. It is much more simple than MySQL InnoDB configuration - you have to set only five parameters: max_connection, shared_buffers, work_mem, maintenance_work_mem and effective_cache_size. Almost all is related to available memory for Postgres on server. Usually work for 5 minutes. On my experience a databases to 100GB is usually without any problems on Postgres (probably on MySQL too). There are two important factors - how speed you expect and how much memory and how fast IO you have.
With large databases you have to have a experience and knowledges for any database technology. All is fast when you are in memory, and when ratio database size/memory is higher, then much more work you have to do to get good results.
First of all, MySQLs table size is only limited by the allowed file size limit of your OS which is I. The terra bytes on any modern OS. That would pose no problems. Most important are questions like this:
What kind of queries will you run?
Are the large table records updated frequently or basically archives for history data?
What is your hardware budget?
What is the kind of query speed you need?
Are you familiar with table partitioning, archive tables, config tuning?
How fast do you need to write (expected inserts per second)
What language will you use to connect to the db (Java, .net, Ruby etc)
What platform are you most familiar with?
Will you run queries which might cause table scans such like '%something%' which would have to go through every single row and take forever
MySQL is used by Facebook, google, twitter and others with large tables and 100,000,000 is not much in the age of social media. MySQL has very little drawbacks (even though I prefer postgresql in most cases) like altering large tables by adding a new index for example. That might send your company in a couple days forced vacation if you don't have a replica in the meantime. Is there a reason why NoSQL is not an option? Sometimes hybrid approaches are a good choice like having your relational business logic in MySQL and huge statistical tables in a NoSQL database like MongoDb which can scale by adding new servers in minutes (MySQL can too but it's more complicated). Now MongoDB can have a indexed column which can be searched by in blistering speed.
Bejond the bottom line: you need to answer the above questions first to make a very informed decision. If you have huge tables and only search on indexed keys almost any database will do - if you expect many changes to the structure down the road you want to use a different approach.
Edit:
Based on your update you just posted I doubt you would run into problems.
I have been recently looking at nosql solutions for our quite big upcoming database and found that cassandra is good but there are very less resources available online about new releases of cassandra and most of the blogs and articles are related to 0.6 version while now it has also implemented support for hadoop and hive. While on the other hand mysql cluster version is also specifically made to run on horizontal scaled setup using commodity servers.
As we are used to relational model for years and moving to cassandra will need decompiling of brain while the product is still not very mature and community is not also that big to respond quickly to any particular problem I have checked datastax(on of the professional support providers) website and their forums are pretty much dead.
So, how to compare mysql cluster vs cassandra while putting relational and non-relational comparison put aside?
Though cassandra is schema less but still it provies pretty much tabular features like super colum and sub column too so record can be searched from multiple column values.
I have also tried my best to find out how cassandra physically stores updated queries like for a row when a sub column is edited and added quite a big chunk of data then how it physically stores that record and how it accesses that record fast? Because in mysql columns have fixed length allocated so its not a big issue.
Here are some areas where I suspect Cassandra has an advantage:
Excellent support for larger-than-memory data sets
Replication: Cassandra supports arbitrary numbers of fully-distributed replicas instead of just partitioned replicas (so, you don't have to have a number of nodes divisible by your replica count in Cassandra, and there are no corner cases to deal with around primary failover), best-in-class support for multiple datacenters, support for synchronous replication as well as asynchronous (important if you're concerned about full durability), and robust self-healing (hinted handoff, read repair, anti-entropy) to make sure you never have to blow away a backup replica and rebuild it from scratch
No locking during ALTER TABLE, index creation, etc
Substantially simpler and less error-prone administration (compare http://dev.mysql.com/doc/refman/5.1/en/mysql-cluster-online-add-node.html and http://wiki.apache.org/cassandra/Operations#Bootstrap). In particular, I would call your attention to how many client or other nodes need to be restarted in the Cassandra scenario: none.
To elaborate on the last a little, most people who haven't actually run Cassandra on a multi-node cluster, don't realize just how well Cassandra has been designed for this. For a two minute taste, see Jake Luciani's demo.
To answer your physical storage question, the key feature that makes Cassandra writes fast is that they are append-only. That is, Cassandra only ever writes sequential blocks to disk; it doesn't need to do any slow seeks to random disk locations during a write.
When a column is updated, two things happen: the write is appended to the commit log (for failure recovery), and the in-memory Memtable is updated. Once the Memtable is full, it is flushed out to disk as a new SSTable. Thus, the length of the data doesn't matter, since you're not trying to fit it into a fixed-length disk structure.
SSTables are read-only - you never go back and overwrite an old value on an update, you just write new ones. On a read, Cassandra first looks in the Memtable for the key. If it doesn't find it, Cassandra scans the SSTables in order from newest to oldest and stops when it finds the key. This gives you the most recent value.
There are a few optimizations as well. Each SSTable has an associated Bloom filter for its keys, which is a compact probabilistic index that can produce false positives but never false negatives. If the key is not in the Bloom filter, you can safely skip that SSTable as it is guaranteed not to contain the key, although you may occasionally read an SSTable that you didn't have to.
When you get too many SSTables, they are merged together into a bigger one in a process called compaction. Essentially this does a big merge sort on the SSTables. This lets Cassandra reclaim the space for values that have been overwritten or deleted, and defragment rows that were spread across multiple SSTables.
See http://www.mikeperham.com/2010/03/13/cassandra-internals-writing/ and http://wiki.apache.org/cassandra/MemtableSSTable for more information.
1st a disclaimer - I work as part of the MySQL Cluster product team
If you are looking to Cluster it would be worth starting with the latest 7.2 Development Release which includes new capabilities to significantly enhance JOIN performnce, as well as a new memcached interface, bypassing the SQL layer
http://dev.mysql.com/tech-resources/articles/mysql-cluster-labs-dev-milestone-release.html
If you are familiar already with MySQL, then the following documentation highlights differences between InnoDB and the current GA 7.1 release:
http://dev.mysql.com/doc/refman/5.1/en/mysql-cluster-ndb-innodb-workloads.html
While these don't provide direct comparisons with Cassandra, they do at least provide the latest information on Cluster from which you can base any comparison
Another option these days is relational model in cassandra with playORM and as long as you partition your really really big tables, you can do joins and all the stuff you are familiar with using Scalable SQL like so
#NoSqlQuery(name="findJoinOnNullPartition", query="PARTITIONS p(:partId) select p FROM TABLE as p INNER JOIN p.security as s where s.securityType = :type and p.numShares = :shares"),
NOTE: The TABLE is a Trades table and p.security references the Security table. Trades is partitioned so it can have unlimited partitions and Security table is smaller so it is not partitioned but you can do all the Scalabla SQL with joins you want to.
Hi there at SO,
I would like some ideas/comments on the following from you honorable and venerable bunch.
I have a 100M records which I need to process. I have 5 nodes (in a rocks cluster) to do this. The data is very structured and falls nicely in the relational data model. I want to do things in parallel since my processing takes some time.
As I see it I have two main options:
Install mysql on each node and put 20M records on each. Use the head node to delegate queries to the nodes and aggregate the results. Query Capabilities++, but I might risk some headaches when I come to choose partitioning strategies etc. (Q. Is this what they call mysql/postgres cluster?). The really bad part is that the processing of the records is left up to me now to take care of (how to distribute across machines etc)...
Alternatively install Hadoop, Hive and HBase (note that this might not be the most efficient way to store my data, since HBase is column oriented) and just define the nodes. We write everything in the MapReduce paradigm and, bang, we live happily ever after. The problem here is that we loose the "real time" query capabilities (I know you can use Hive, but that is not suggested for real time queries - which I need) - since I also have some normal sql queries to execute at times "select * from wine where colour = 'brown'".
Note that in theory - if I had 100M machines I could do the whole thing instantly since for each record the processing is independent of the other. Also - my data is read-only. I do not envisage any updates happening. I do not need/want 100M records on one node. I do not want there to be redundant data (since there is lots of it) so keeping it in BOTH mysql/postgres and Hadoop/HBase/HDFS. is not a real option.
Many Thanks
Can you prove that MySQL is the bottleneck? 100M records is not that many, and it looks like that you're not performing complex queries. Without knowing exactly what kind of processing, here is what I would do, in this order:
Keep the 100M in MySQL. Take a look at Cloudera's Sqoop utility to import records from the database and process them in Hadoop.
If MySQL is the bottleneck in (1), consider setting up slave replication, which will let you parallelize reads, without the complexity of a sharded database. Since you've already stated that you don't need to write back to the database, this should be a viable solution. You can replicate your data to as many servers as needed.
If you are running complex select queries from the database, and (2) is still not viable, then consider using Sqoop to import your records and do whatever query transformations you require in Hadoop.
In your situation, I would resist the temptation to jump off of MySQL, unless it is absolutely necessary.
There are a few questions to ask, before suggesting.
Can you formulate your queries to access by primary key only? In other words - if you can avoid all joins and table scans. If so - HBase is an option, if you need very high rate of read/write accesses.
I do noth thing that Hive is good option taking into consideration low data volume. If you expect them to grow significantly - you can consider it. In any case Hive is good for the analytical workloads - not for the OLTP type of processing.
If you do need relational model with joins and scans - I think good solution might be one Master Node and 4 slaves, with replication between them. You will direct all writes to the master, and balance reads among whole cluster. It is especially good if you have much more reads then writes.
In this schema you will have all 100M records (not that match) on each node. Within each node you can employ partitioning if appropriate.
You may also want to consider using Cassandra. I recently discovered this article on HBase vs. Cassandra which I was reminded of when I read your post.
The gist of it is that Cassandra is a highly scallable NoSQL solution with fast querying, which sort of sounds like the solution you're looking for.
So, it all depends on whether you need to maintain your relational model or not.
HI,
I had a situation where I had many tables which I created in parallel using sqlalchemy and the python multiprocessing library. I had multiple files, one per table, and loaded them using parallel COPY processes. If each process corresponds to a separate table, that works well. With one table, using COPY would be difficult. You could use tables partitioning in PostgreSQL, I guess. If you are interested I can give more details.
Regards.