For a customer I am currently investigating improvements to their database structure.
My customer offers holiday rentals on their website.
On their front page they have a search function wich sends a query to a MySQL database architecture (Master-Master setup) that answers that query with all the holiday rentals that the customer is interested in.
Due to the growth of the company and the increasing load on their servers the search query's are currently running up to 10+ seconds. Mainly because the query's end with an ORDER BY which causes MySQL to create a temp table and sort all the data, an average search query can return up to 20k holiday homes.
Ofcourse one of the things we are doing is investigating the query's, rewriting them and putting indexes where needed. Unfortunately we are unable to get allot more performance under these circumstances.
That's why we are looking into implementing Memcached on top of MySQL to cache these large datasets in memory for faster retrieval. Unfortunately the datasets that the query's return are quite large wich makes Memcached not that effective at this point. The array that MySQL returns are currently about 15k rows with about 60 values per row.
The reason Memcached is interesting is because we want to drastically improve the search function, and lowering the load on the MySQL platform. This would make it more scalable.
I am wondering if there is anyone that is familair with (longterm) caching MySQL data in Memcached and making it more effective for large datasets?
Thanks a bunch!
Memcache is for storing key-value pairs, not for large sets of data. Will it work? Yes. Of course it will. But with how much data you guys are going to throw at it, you're going to run out of memory very soon and end up hitting the database anyway with how often your search results may change. And remember that just because it's memcache doesn't mean it doesn't have to go through web sockets to a (most likely) different machine. Your problem seems to be that you're using MySQL for something it was never designed well for, which is its use as a search engine. No matter how many things you optimize, all you're doing is raising the ceiling an inch at a time.
I could take this post in a "you need to optimize MySQL parameters so that it doesn't have to create those temp tables" direction, but I'm going to assume you've already looked into that and keep going.
My recommendation is that you implement something on top of MySQL to handle the searching. In my own quest for fast searching, these are the solutions I gave the most weight to:
Sphinx: http://sphinxsearch.com
Solr: http://lucene.apache.org/solr
Elasticsearch: http://www.elasticsearch.org
You'll find plenty of resources here on StackOverflow for which of those is better and faster and what not. For our purposes, we picked Elasticsearch for one of our projects and Solr for another.
Related
I'm studying up on the future of the database I maintain. Right now we have one database server running MySQL using InnoDB and MyISAM tables. I'm watching the metrics closely and I can see that this will not be sustainable forever. Where does one go next? I have reviewed solutions like Cassandra, but I want to stick to an SQL approach so I'm not sure about that. I have also reviewed NDB cluster and federated database solutions, but I've noticed no one has anything good to say about those. Basically, I looking for advice on intermediate solutions. We do not yet need a vast multi-node array operating on tens of DB servers, but one server is about to reach its limit. I don't want to just throw another server on the pile without making sure that the DB architecture at hand benefits well from the extra power. What do you guys suggest for when it is time to move beyond a single server and how to manage this transition. Thank you to anyone who can help.
Edit to better explain: At present, we have about a hundred tables. We run many join operations to gather the data the end user needs to see, such that most of our queries join at least two tables to complete any operation. The data set is not too big yet, only a few hundred Megs, but the data is accessed in such a way that each table has a few writes everyday, the heaviest of which has about a thousand writes a day. We probably have about a few hundred thousand reads a day too, so read do outnumber writes about 9 to 1.
First Solutions:
Indices go a LONG way
Use profiling software to find your slow queries and optimize them
Depending on your hosting company you can usually update the RAM/CPU of the server
Second Solutions:
Split your reads and your writes into two databases. (I don't know if you're using PHP or not but PHP has a plugin that will automatically split them for you without having to change any of your code http://php.net/manual/en/mysqlnd-ms.rwsplit.php)
Use software like memcache to store database information that is frequently queried but not frequently updated
In our (currently MySQL) database there are over 120 million records, and we make frequent use of complex JOIN queries and application-level logic in PHP that touch the database. We're a marketing company that does data mining as our primary focus, so we have many large reports that need to be run on a daily, weekly, or monthly basis.
Concurrently, customer service operates on a replicated slave of the same database.
We would love to be able to make these reports happen in real time on the web instead of having to manually generate spreadsheets for them. However, many of our reports take a significant amount of time to pull data for (in some cases, over an hour).
We do not operate in the cloud, choosing instead to operate using two physical servers in our server room.
Given all this, what is our best option for a database?
I think you're going the wrong way about the problem.
Thinking if you drop in NoSQL that you'll get better performance is not really true. At the lowest level, you're writing and retrieving a fair chunk of data. That implies your bottleneck is (most likely) HDD I/O (which is the common bottleneck).
Sticking to the hardware you have momentarily and using a monolithic data storage isn't scalable and as you noticed - has implications when wanting to do something in real-time.
What are your options? You need to scale your server and software setup (which is what you'd have to do with any NoSQL anyway, stick in faster hard drives at some point).
You also might want to look into alternative storage engines (other than MyISAM and InnoDB - for example, one of better engines that seemingly turn random I/O to sequential I/O is TokuDB).
Implementing faster HDD subsystem would also aid to your needs (FusionIO if you have the resources to get it).
Without more information on your end (what the server setup is, what MySQL version you're using and what storage engines + data sizes you're operating with), it's all speculation.
Cassandra still needs Hadoop for MapReduce, and MongoDB has limited concurrency with regard to MapReduce...
... so ...
... 120 mio records is not that much, and MySQL should easily be able to handle that. My guess is an IO bottleneck, or you're doing lots of random reads instead of sequential reads. I'd rather hire a MySQL techie for a month or so to tune your schema and queries, instead of investing into a new solution.
If you provide more information about your cluster, we might be able to help you better. "NoSQL" by itself is not the solution to your problem.
As much as I'm not a fan of MySQL once your data gets large, I have to say that you're nowhere near needing to move to a NoSQL solution. 120M rows is not a big deal: the database I'm currently working with has ~600M in one table alone and we query it efficiently. Managing that much data from an ops perspective is the problem; querying it isn't.
It's all about proper indexes and the correct use of them when joining, and secondarily memory settings. Find your slow queries (mysql slow query log FTW!), and learn to use the explain keyword to understand whey they are slow. Then tweak your indexes so your queries are efficient. Further, make sure you understand MySQL's memory settings. There are great pages in the docs explaining how they work, and they aren't that hard to understand.
If you've done both of those things and you're still having problems, make sure disk I/O isn't an issue. Then you should look in to another solution for querying your data if it is.
NoSQL solutions like Cassandra have a lot of benefits. Cassandra is fantastic at writing data. Scaling your writes is very easy--just add more nodes! But the tradeoff is that it's harder to get the data back out. From a cost perspective, if you have expertise in MySQl, it's probably better to leverage that and scale your current solution until it hits a limit before completely switching your underlying architecture.
I am about 70% of the way through developing a web application which contains what is essentially a largeish datatable of around 50,000 rows.
The app itself is a filtering app providing various different ways of filtering this table such as range filtering by number, drag and drop filtering that ultimately performs regexp filtering, live text searching and i could go on and on.
Due to this I coded my MySQL queries in a modular fashion so that the actual query itself is put together dynamically dependant on the type of filtering happening.
At the moment each filtering action (in total) takes between 250-350ms on average. For example:-
The user grabs one end of a visual slider, drags it inwards, when he/she lets go a range filtering query is dynamically put together by my PHP code and the results are returned as a JSON response. The total time from the user letting go of the slider until the user has recieved all data and the table is redrawn is between 250-350ms on average.
I am concerned with scaleability further down the line as users can be expected to perform a huge number of the filtering actions in a short space of time in order to retrieve the data they are looking for.
I have toyed with trying to do some fancy cache expiry work with memcached but couldn't get it to play ball correctly with my dynamically generated queries. Although everything would cache correctly I was having trouble expiring the cache when the query changes and keeping the data relevent. I am however extremely inexperienced with memcached. My first few attempts have led me to believe that memcached isn't the right tool for this job (due to the highly dynamic nature of the queries. Although this app could ultimately see very high concurrent usage.
So... My question really is, are there any caching mechanisms/layers that I can add to this sort of application that would reduce hits on the server? Bearing in mind the dynamic queries.
Or... If memcached is the best tool for the job, and I am missing a piece of the puzzle with my early attempts, can you provide some information or guidance on using memcached with an application of this sort?
Huge thanks to all who respond.
EDIT: I should mention that the database is MySQL. The siite itself is running on Apache with an nginx proxy. But this question is related purely to speeding up and reducing the database hits, of which there are many.
I should also add that the quoted 250-350ms roundtrip time is fully remote. As in from a remote computer accessing the website. The time includes DNS lookup, Data retrieval etc.
If I understand your question correctly, you're essentially asking for a way to reduce the number of queries against the database eventhough there will be very few exactly the same queries.
You essentially have three choices:
Live with having a large amount of queries against your database, optimise the database with appropriate indexes and normalise the data as far as you can. Make sure to avoid normal performance pitfalls in your query building (lots of ORs in ON-clauses or WHERE-clauses for instance). Provide views for mashup queries, etc.
Cache the generic queries in memcached or similar, that is, without some or all filters. And apply the filters in the application layer.
Implement a search index server, like SOLR.
I would recommend you do the first though. A roundtrip time of 250~300 ms sounds a bit high even for complex queries and it sounds like you have a lot to gain by just improving what you already have at this stage.
For much higher workloads, I'd suggest solution number 3, it will help you achieve what you are trying to do while being a champ at handling lots of different queries.
Use Memcache and set the key to be the filtering query or some unique key based on the filter. Ideally you would write your application to expire the key as new data is added.
You can only make good use of caches when you occasionally run the same query.
A good way to work with memcache caches is to define a key that matches the function that calls it. For example, if the model named UserModel has a method getUser($userID), you could cache all users as USER_id. For more advanced functions (Model2::largerFunction($arg1, $arg2)) you can simply use MODEL2_arg1_arg2 - this will make it easy to avoid namespace conflicts.
For fulltext searches, use a search indexer such as Sphinx or Apache Lucene. They improve your queries a LOT (I was able to do a fulltext search on a 10 million record table on a 1.6 GHz atom processor, in less than 500 ms).
I am working with large datasets (10s of millions of records, at times, 100s of millions), and want to use a database program that links well with R. I am trying to decide between mysql and sqlite. The data is static, but there are lot of queries that I need to do.
In this link to sqlite help, it states that:
"With the default page size of 1024 bytes, an SQLite database is limited in size to 2 terabytes (241 bytes). And even if it could handle larger databases, SQLite stores the entire database in a single disk file and many filesystems limit the maximum size of files to something less than this. So if you are contemplating databases of this magnitude, you would do well to consider using a client/server database engine that spreads its content across multiple disk files, and perhaps across multiple volumes."
I'm not sure what this means. When I have experimented with mysql and sqlite, it seems that mysql is faster, but I haven't constructed very rigorous speed tests. I'm wondering if mysql is a better choice for me than sqlite due to the size of my dataset. The description above seems to suggest that this might be the case, but my data is no where near 2TB.
I'd appreciate any insights into understanding this constraint of maximum file size from the filesystem and how this could affect speed for indexing tables and running queries. This could really help me in my decision of which database to use for my analysis.
The SQLite database engine stores the entire database into a single file. This may not be very efficient for incredibly large files (SQLite's limit is 2TB, as you've found in the help). In addition, SQLite is limited to one user at a time. If your application is web based or might end up being multi-threaded (like an AsyncTask on Android), mysql is probably the way to go.
Personally, since you've done tests and mysql is faster, I'd just go with mysql. It will be more scalable going into the future and will allow you to do more.
I'm not sure what this means. When I have experimented with mysql and sqlite, it seems that mysql is faster, but I haven't constructed very rigorous speed tests.
The short short version is:
If your app needs to fit on a phone or some other embedded system, use SQLite. That's what it was designed for.
If your app might ever need more than one concurrent connection, do not use SQLite. Use PostgreSQL, MySQL with InnoDB, etc.
It seems that (in R, at least), that SQLite is awesome for ad hoc analysis. With the RSQLite or sqldf packages it is really easy to load data and get started. But for data that you'll use over and over again, it seems to me that MySQL (or SQL Server) is the way to go because it offers a lot more features in terms of modifying your database (e.g., adding or changing keys).
SQL if you are mainly using this as a web service.
SQLite, if you want it to able to function offline.
SQLite generally is much much faster, as majority (or ALL) of data/indexes will be cached in memory. However, in the case of SQLite. If the data is split up across multiple tables, or even multiple SQLite database files, from my experience so far. For even millions of records (i yet to have 100's of millions though), it is far more effective then SQL (compensate the latency / etc). However that is when the records are split apart in differant tables, and queries are specific to such tables (dun query all tables).
An example would be a item database used in a simple game. While this may not sound much, a UID would be issued for even variations. So the generator soon quickly work out to more then a million set of 'stats' with variations. However this was mainly due to each 1000 sets of records being split among different tables. (as we mainly pull records via its UID). Though the performance of splitting was not properly measured. We were getting queries that were easily 10 times faster then SQL (Mainly due to network latency).
Amusingly though, we ended up reducing the database to a few 1000 entries, having item [pre-fix] / [suf-fix] determine the variations. (Like diablo, only that it was hidden). Which proved to be much faster at the end of the day.
On a side note though, my case was mainly due to the queries being lined up one after another (waiting for the one before it). If however, you are able to do multiple connections / queries to the server at the same time. The performance drop in SQL, is more then compensated, from your client side. Assuming this queries do not branch / interact with one another (eg. if got result query this, else that)
I'm wondering if some other non-relational database would be a good fit for activity streams - sort of like what you see on Facebook, Flickr (http://www.flickr.com/activity), etc. Right now, I'm using MySQL but it's pretty taxing (I have tens of millions of activity records) and since they are basically read-only once written and always viewed chronologically, I was thinking that an alternative DB might work well.
The activities are things like:
6 PM: John favorited Bacon
5:30 PM: Jane commented on Snow Crash
5:15 PM: Jane added a photo of Bacon to her album
The catch is that unlike Twitter and some other systems, I can't just simply append activities to lists for each user who is interested in the activity - if I could it looks like Redis would be a good fit (with its list operations).
I need to be able to do the following:
Pull activities for a set or subset of people who you are following ("John" and "Jane"), in reverse date order
Pull activities for a thing (like "Bacon") in reverse date order
Filter by activity type ("favorite", "comment")
Store at least 30 million activities
Ideally, if you added or removed a person who you are following, your activity stream would reflect the change.
I have been doing this with MySQL. My "activities" table is as compact as I could make it, the keys are as small as possible, and the it is indexed appropriately. It works, but it just feels like the wrong tool for this job.
Is anybody doing anything like this outside of a traditional RDBMS?
Update November 2009: It's too early to answer my own question, but my current solution is to stick with MySQL but augment with Redis for fast access to the fresh activity stream data. More information in my answer here: How to implement the activity stream in a social network...
Update August 2014: Years later, I'm still using MySQL as the system of record and using Redis for very fast access to the most recent activities for each user. Dealing with schema changes on a massive MySQL table has become a non-issue thanks to pt-online-schema-change
I'd really, really, suggest stay with MySQL (or a RDBMS) until you fully understand the situation.
I have no idea how much performance or much data you plan on using, but 30M rows is not very many.
If you need to optimise certain range scans, you can do this with (for example) InnoDB by choosing a (implicitly clustered) primary key judiciously, and/or denormalising where necessary.
But like most things, make it work first, then fix performance problems you detect in your performance test lab on production-grade hardware.
EDIT:Some other points:
key/value database such as Cassandra, Voldermort etc, do not generally support secondary indexes
Therefore, you cannot do a CREATE INDEX
Most of them also don't do range scans (even on the main index) because they're using hashing to implement partitioning (which they mostly do).
Therefore they also don't do range expiry (DELETE FROM tbl WHERE ts < NOW() - INTERVAL 30 DAYS)
Your application must do ALL of this itself or manage without it; secondary indexes are really the killer
ALTER TABLE ... ADD INDEX takes quite a long time in e.g. MySQL with a large table, but at least you don't have to write much code to do it. In a "nosql" database, it will also take a long time BUT also you have to write heaps and heaps of code to maintain the new secondary index, expire it correctly, AND modify your queries to use it.
In short... you can't use a key/value database as a shortcut to avoid ALTER TABLE.
I am also planning on moving away from SQL. I have been looking at CouchDB, which looks promising. Looking at your requirements, I think all can be done with CouchDB views, and the list api.
It seems to me that what you want to do -- Query a large set of data in several different ways and order the results -- is exactly and precisely what RDBMeS were designed for.
I doubt you would find any other datastore that would do this as well as a modern commercial DBMS (Oracle, SQLServer, DB2 etc.) or any opn source tool that would accomplish
this any better than MySql.
You could have a look at Googles BigTable, which is really a relational database but
it can present an 'object'y personality to your program. Its exceptionaly good for free format text
searches, and complex predicates. As the whole thing (at least the version you can download) is implemented in Python I doubt it would beat MySql in a query marathon.
For a project I once needed a simple database that was fast at doing lookups and which would do lots of lookups and just an occasional write. I just ended up writing my own file format.
While you could do this too, it is pretty complex, especially if you need to support it from a web server. With a web server, you would at least need to protect every write to the file and make sure it can be read from multiple threads. The design of this file format is something you should work out as good as possible with plenty of testing and experiments. One minor bug could prove fatal for a web project in this style, but if you get it working, it can work real well and extremely fast.
But for 99.999% of all situations, you don't want such a custom solution. It's easier to just upgrade the hardware, move to Oracle, SQL Server or InterBase, use a dedicated database server, use faster hard disks, install more memory, upgrade to a 64-bit system. Those are the more generic tricks to improve performance with the least effort.
I'd recommend learning about message queue technology. There are several open-source options available, and also robust commercial products that would serve up the volume you describe as a tiny snack.
CouchDB is schema-free, and it's fairly simple to retrieve a huge amount of data quickly, because you are working only with indexes. You are not "querying" the database each time, you are retrieving only matching keys (which are pre-sorted making it even faster).
"Views" are re-indexed everytime new data is entered into the database, but this takes place transparently to the user, so while there might be potential delay in generating an updated view, there will virtually never be any delay in retrieving results.
I've just started to explore building an "activity stream" solution using CouchDB, and because the paradigm is different, my thinking about the process had to change from the SQL thinking.
Rather than figure out how to query the data I want and then process it on the page, I instead generate a view that keys all documents by date, so I can easily create multiple groups of data, just by using the appropriate date key, essentially running several queries simultaneously, but with no degradation in performance.
This is ideal for activity streams, and I can isolate everything by date, or along with date isolation I can further filter results of a particular subtype, etc - by creating a view as needed, and because the view itself is just using javascript and all data in CouchDB is JSON, virtually everything can be done client-side to render your page.