Give Your Site A Boost With Memcache (OSCON 2008)

Today’s high-traffic web sites must implement performance-boosting measures that cut down data processing and reduce load on the database, while increasing the speed of content delivery. One such method is the use of a cache to temporarily store whole pages, database recordsets, large objects, and sessions. While many caching mechanisms exist, memcached provides one of the fastest and easiest-to-use caching servers. This talk will cover memcached and the memcache extension for PHP from setting up a memcached server to using it to provide a variety of caching solutions, including the use of memcached as a session data store.

While all examples will be given in PHP, the caching principles illustrated using memcached should also apply to other programming languages.

Transcript

1. Ben Ramsey
   July 23, 2008
   Give Your Site A Boost With Memcache
   

   View Slide

2. Why cache?
   2
   

   View Slide

3. To make it faster.
   3
   

   View Slide

4. 4
   “A cache is a collection of data duplicating original
   values stored elsewhere or computed earlier,
   where the original data is expensive to fetch
   (owing to longer access time) or to compute,
   compared to the cost of reading the cache. In
   other words, a cache is a temporary storage area
   where frequently accessed data can be stored for
   rapid access.”
   — Wikipedia
   

   View Slide

5. Why cache?
   5
   You want to reduce the number of retrieval
   queries made to the database
   You want to reduce the number of external
   requests (retrieving data from other web
   services)
   You want to cut down on filesystem access
   

   View Slide

6. Caching options
   6
   Flat file caching
   Caching data in the database
   MySQL 4.x query caching
   Shared memory (APC)
   RAM disk
   memcached
   

   View Slide

7. What is memcached?
   7
   Distributed Memory Object Caching System
   Caching daemon
   Developed by Danga Interactive for
   LiveJournal.com
   Uses RAM for storage
   Acts as a dictionary of stored data with key/
   value pairs
   

   View Slide

8. Is memcached fast?
   8
   Stored in memory (RAM), not on disk
   Uses non-blocking network I/O (TCP/IP)
   Uses libevent to scale to any number of open
   connections
   Uses its own slab allocator and hash table to
   ensure virtual memory never gets externally
   fragmented and allocations are guaranteed
   O(1)
   

   View Slide

9. General usage
   9
   1.Set up a pool of memcached servers
   2.Assign values to keys that are stored in the
   cluster
   3.The memcache client hashes the key to a
   particular machine in the cluster
   4.Subsequent requests for that key retrieve the
   value from the memcached server on which it
   was stored
   5.Values time out after the specified TTL
   

   View Slide

10. Memcached principles
    10
    It’s a non-blocking server
    It is not a database
    It does not provide redundancy
    It doesn't handle failover
    It does not provide authentication
    

    View Slide

11. Memcached principles
    11
    Data is not replicated across the cluster
    Works great on a small and local-area
    network
    A single value cannot contain more than 1MB
    of data
    Keys are strings limited to 250 characters
    

    View Slide

12. Storing data in the pool
    12
    Advantage is in scalability
    To fully see the advantage, use a “pool”
    memcached itself doesn't know about the pool
    The pool is created by and managed from the
    client library
    

    View Slide

13. www 2
    memcached
    www 1
    www 3
    memcached
    memcached
    13
    

    View Slide

14. Deterministic failover
    14
    When one server goes down, the system fails
    over to another server in the pool
    Memcached does not provide this
    Some memcache clients provide failover
    If you can’t find the data in memcache, eat the
    look-up cost and retrieve from your data
    source again, storing it back to the cache
    

    View Slide

15. www 2
    memcached
    www 1
    www 3
    memcached
    memcached
    www 3
    memcached
    Data
    inaccessible!
    Recreate data;
    Store back to
    memcache
    15
    

    View Slide

16. The memcached protocol API
    16
    Storage commands:
    set, add, replace, append, prepend, cas
    Retrieval command: get, gets
    Deletion command: delete
    Increment/decrement: incr, decr
    Other commands:
    stats, flush_all, version, verbosity,
    quit
    

    View Slide

17. $> telnet localhost 11211
    Trying ::1...
    Connected to localhost.
    Escape character is '^]'.
    set foobar 0 0 15
    This is a test.
    STORED
    get foobar
    VALUE foobar 0 15
    This is a test.
    END
    quit
    Connection closed by foreign host.
    $>
    17
    

    View Slide

18. Setting it up
    18
    http://danga.com/memcached/
    $> ./configure; make; make install
    $> memcached -d -m 2048 -p 11211
    Done!
    Windows port of v1.2.4 at
    http://www.splinedancer.com/memcached-win32/
    

    View Slide

19. Memcached clients
    19
    Perl, Python, Ruby, Java, C#
    C (libmemcached)
    PostgreSQL (access memcached from procs
    and triggers)
    MySQL (adds memcache_engine storage
    engine)
    PHP (pecl/memcache)
    

    View Slide

20. pecl/memcache
    20
    The PHP client for connecting to memcached
    and managing a pool of memcached servers
    http://pecl.php.net/package/memcache
    $> pecl install memcache
    Stable: 2.2.3
    Beta: 3.0.1
    

    View Slide

21. View Slide

22. Features of pecl/memcache
    22
    memcache.allow_failover
    memcache.hash_strategy
    memcache.hash_function
    memcache.protocol
    memcache.redundancy
    memcache.session_redundancy
    

    View Slide

23. pecl/memcache interface
    MemcachePool::connect()
    MemcachePool::addServer()
    MemcachePool::setServerParams()
    MemcachePool::get()
    MemcachePool::add()
    MemcachePool::set()
    MemcachePool::replace()
    MemcachePool::cas()
    23
    

    View Slide

24. pecl/memcache interface
    MemcachePool::append()
    MemcachePool::prepend()
    MemcachePool::delete()
    MemcachePool::increment()
    MemcachePool::decrement()
    MemcachePool::setFailureCallback()
    24
    

    View Slide

25. Key hashing
    25
    Keys longer than 250 characters are
    truncated without warning
    Good practice to hash your key (with MD5 or
    SHA) at the userland level to ensure long
    keys don’t get truncated
    Keys are “global”
    Use something to uniquely identify keys, e.g.
    a method signature or an SQL statement
    

    View Slide

26. Object serialization
    26
    Objects are serialized before being stored to
    memcache:
    get key
    VALUE key 1 59
    O:8:"stdClass":2:{s:3:"foo";s:3:"bar";s:
    3:"baz";s:3:"quz";}
    END
    Extension unserializes them before returning
    the object
    Only objects that can be serialized safely can
    be stored to memcache, i.e. problems with
    DOM, SimpleXML, etc.
    

    View Slide

27. Redundancy and failover
    27
    memcache.redundancy &
    memcache.session_redundancy
    Implement redundancy at the userland level?
    Again, memcache is not a database
    

    View Slide

28. Extending MemcachePool
    28
    Implement global values vs. page-specific
    values
    Ensure a single instance of the
    MemcachePool object
    Do complex key hashing, if you so choose
    Set a default expiration for all your data
    Add all of your servers upon object
    instantiation
    

    View Slide

29. Database techniques
    29
    Create a wrapper for mysql_query() that
    checks the cache first and returns an array of
    database results
    Extend PDO to store results to the cache and
    get them when you execute a statement
    

    View Slide

30. Database techniques
    30
    For large datasets, run a scheduled query
    once an hour and store it to the cache
    Please note: memcached can store arrays,
    objects, etc., but it cannot store a resource,
    which some database functions (e.g.
    mysql_query()) return
    

    View Slide

31. Session storage
    31
    As of 2.1.1, you can set the session save
    handler as “memcache” and all will work
    automagically
    session.save_handler = memcache
    session.save_path = "tcp://
    192.168.1.10:11211,tcp://
    192.168.1.11:11211,tcp://192.168.1.12:11211"
    Store sessions to both the database and
    memcache
    Write your own session handler that stores to
    the database and memcache
    

    View Slide

32. www 3
    memcached
    www 2
    memcached
    www 1
    memcached
    Session
    inaccessible!
    Need to
    recreate the
    session!
    32
    

    View Slide

33. 33
    A demonstration...
    

    View Slide

34. 34
    For more information...
    http://danga.com/memcached/
    http://pecl.php.net/package/memcache
    http://www.socialtext.net/memcached/
    

    View Slide

35. Thank You
    Slides available for download at benramsey.com.
    Ben Ramsey
    Software Architect
    Schematic
    http://www.schematic.com/
    35
    

    View Slide

36. We’re Hiring!
    It goes without saying: Schematic is only as good
    as the people who work here. That’s why we’re so
    particular about recruiting, training, nurturing, and
    retaining the very best people in our field.
    If you have digital expertise (technical, creative, managerial–or
    something else entirely), enthusiasm, curiosity, and the ability
    to collaborate with others, we’d love to hear from you.
    Please visit http://www.schematic.com/ for more information.
    36
    

    View Slide