MemCached Non-Traditional Usage?

Travis Boucher tbone at
Wed Mar 21 15:23:10 UTC 2007

The memcached clients hash the keys among the list of instances.  As 
long as all of the clients are using the same hashing and list of 
memcached instances, then they will map the same keys to the same servers.

Geoffrey Mina wrote:
> Thank for all your your responses!!  I never expected this kind of 
> response from a smaller group... truly awesome!
> Perhaps I am missing something in the documentation... or maybe I am 
> missing the bigger picture.  If I have 3-5 memcached servers how do my 
> clients know which server to retrieve the object from?  As far as I 
> can tell the data is not distributed between the memcached 
> instances... so you _must_ request the object from the same memcached 
> server it was originally stored in.  The server which writes the data 
> will not always be the server that needs to read the data...
> To the fellow who suggested MySQL cluster, I am trying to stay away 
> from a database solution for this problem.  Our architecture already 
> has a lot of database nodes.  We use a MySQL shard model where the 
> load is balanced logically by account/customer.  All master 
> servers are running DRBD, with a global catalog replicated locally, 
> etc, etc, etc... needless to say it is already a LOT of database 
> servers to manage... The data I am hoping to store in memcached is 
> simply routing information so the processes know which MySQL shard 
> should be used for reading/writing data. 
> Thanks again!
> Geoff
> On 3/21/07, *Dave Strickler* <dstrickler at 
> <mailto:dstrickler at>> wrote:
>     Your usage will work just fine. Along with it's traditional use,
>     we do the same thing in house. As long as you know the key for the
>     data you're looking for I think you'll be very happy with the results.
>     Dave Strickler
>     MailWise LLC
>     617-933-5810 (direct)
> <>
>     "Intelligent E-mail Protection"
>     >>> "Geoffrey Mina" <geoffreymina at
>     <mailto:geoffreymina at>> 11:40 PM Tuesday, March 20, 2007 >>>
>     Hello,
>     I am currently evaluating memcached for a non-traditional
>     application.  It seems like the core functionality of memcached is
>     to help reduce the number of database reads that are required for
>     read-heavy web applications.  I am looking to implement memcached
>     in totally different environment, and wondering if I am
>     understanding the software correctly.
>     I am hoping to implement this as a solution for a telephony
>     platform.  I have a cluster of servers which need to share
>     information with each other.   I am looking to store information
>     into memcached with server 1 and retrieve it with server 2... no
>     database involved at any point.  The data that would be existing
>     in cache would be true transient data that would exist for seconds
>     or minutes and then would be flushed from the shared memory/cache. 
>     If I have 3 memcached servers and 10 telephony servers running the
>     client API (connecting to all 3), would they freely share
>     information with each other... or do I have this whole thing wrong?
>     Thanks In Advance!
>     Geoff

More information about the memcached mailing list