proper memcached usage

[Marc Rossi]

I am planning on using a 0 expiration and purging the cache nightly with a memcached restart.  The consumer apps have been written in such a way that they will warn and wait for the data to be available (at least that is how they are in the postgres version).  Unfortunately the setup for accessing the data feed directly in all apps is just to much overhead to incur.


----- Original Message ----
From: Dustin Sallings <dustin at spy.net>
To: Marc Rossi <marc_rossi at yahoo.com>
Cc: memcached at lists.danga.com
Sent: Tuesday, September 11, 2007 4:23:23 PM
Subject: Re: proper memcached usage


On Sep 11, 2007, at 14:13 , Marc Rossi wrote:

> I'm looking to replace a local machine data distribution system  
> using postgres with memcached.  There will be one process that  
> updates approximately 1500 items every second.  Each item will  
> point to a comma separated list of floats (10 max).  On the same  
> machine there will be several processes that consume this data in  
> different intervals.
>
> Is this a suitable use for memcached?  Most of the examples I have  
> read about involve multiple writers & readers, in this scenario I  
> will only have 1 writer and multiple readers (probably a max of 10).

    The biggest issue is that you can't rely on values being there when  
you go back to read them.

    Memcached only guarantees the values to *not* be there after  
whatever expiration date you set.  It doesn't make any guarantees  
about when the values *will* be there.

-- 
Dustin Sallings