Using memcached with ASP.NET

Posted by Jonathan Allen on Jul 12, 2007

Sections

Development,

Architecture & Design

Topics

Clustering & Caching ,

.NET

Tags

Memcache

Instead of ASP.Net's built-in caching, some .NET developers are turning to memcached, is a distributed memory caching system originally by Danga Interactive for LiveJournal.

A fundamental problem with caching is stale data. When running a single web server, one can easily clear a cache when data is known to have been changed. Unfortunately, ASP.NET doesn't have a good way to scale this up multiple servers. Each server's cache is blissfully unaware of changes to other caches.

ASP.NET does allow triggers based on changes to the file system or a database table to invalidate a cache. However these have problems, such as the expensive polling that database triggers cause and the tedious wiring of the triggers themselves. There are other options however.

Unlike ASP.NET's built-in caching, memcached is a distributed cache. Any web server in can update or delete a cache entry and all the other servers automatically see the change next time they access the cache. This is done by storing the entries on one or more cache servers. Each entry assigned to a server based on a hash of its key.

Superficially, the ASP.NET API for memcached looks almost identical to the built-in API. This allows switching to memcached to be as easy as a single pass with search and replace.

Moving beyond just getting it running however, there are some questions as its proper use in larger web farms. Richard Jones writes

As we add more nodes, the usefulness of get_multi decreases - it's possible for a single page to hit almost all of the memcached instances. I read somewhere that facebook partition their memcached cluster to improve get_multi performance (eg, all user data on a subset of mc nodes). Can anyone comment on the effectiveness of this?

One proposed solution is to generate hash keys separately from the cache entire's key. This would allow a developer to ensure that all entries needed by a given page are more likely to be on the same server. Unfortunately, generating hash keys based on where one wants the data stored rather than from the cache key itself is likely to be error prone and will take careful implementation.

You can download memcached under a BSD license. Client APIs for C#, as well as Perl, Python, PHP, Java, and a host of other languages have to installed separately. Finally, there is a Win32 port for those not wanting to run Linux machines.

 

1. Back to top

   Great Stuff

   by Aaron Erickson

   I am actually doing some work on a side project that could very much use this! Thanks for the story.

   Reply

2. Back to top

   Re: Great Stuff

   by Nati Shalom

   You can find similar and much more enhanced implementation of caching solution A.K.A DataGrid through GigaSpaces
   
   It comes with built-in partitioning, SLA driven container, support for Spring, SQL query support etc. A totally free community edition is also available which support basic hub/spoke clustering, persistencey etc.
   
   
   And there is a .net implementation available as well.
   
   HTH
   Nati S.
   
   GigaSpaces write once scale anywhere

   Reply

3. Back to top

   Coherence

   by Cameron Purdy

   We've been seeing a good number of migrations recently from Memcached to Coherence lately. In addition to the support model, the main reason seems to be the issues with the "opaqueness" of using Memcached, plus necessary features such as reliable write-behind, which are necessary for high performance database integration.
   
   Peace,
   
   Cameron Purdy
   Oracle Coherence: Data Grid

   Reply

4. Back to top

   Re: CSQL Cache

   by kanchana p

   If you are looking for an open source alternative, CSQL Cache is the answer.
   It support bi-directional, updateable, real time table caching for applications. Check out this blog for its functionalites, csqlcache.wordpress.com
   
   Project Page:
   sourceforge.net/projects/csql

   Reply

• Older >