PECL memcache extension

Don MacAskill don at
Tue Feb 7 17:23:26 UTC 2006

Let's be really clear here:

1 - All of my data is in a RDBMS.  We're not talking about losing any 
data when memcached (or the server or network on which it resides) dies. 
  I'm just interested in preventing stale reads of that data and in 
preventing lots of reads off (slow) disks.  I want it to exactly match 
the data in my RDBMS at all times.  That's not difficult to do, and it's 
massively more scalable than a disk-based RDBMS.

2 - I have this working *today*.  Again, it's not very difficult.  The 
only reason for the thread is to see if there's interest in extending 
the PECL extension to support this sort of scenario.  I happen to think 
it's very useful - you get all of the speed of memcached with next to no 
worry about stale reads, even in multi-server failure scenarios.  What's 
not to like?

Like I said earlier, I'm happy to share my code so people can see how 
simple it is, but I want to make sure I understand whether the PECL 
extension can/should/will be extended in this fashion or not before I 
try to adapt our current mechanism to use the extension instead of our 
homebrewed php-client modification.


Brian Moon wrote:
>> Of course, all of this depends on how much of a problem it is when you 
>> get stale data.  I guess in any situation where it _really_ matters, 
>> you shouldn't be using memcached in the first place, but something 
>> more persistent / less error prone such as db transactions.
> I have been thinking the same thing through all this.  It sounds like 
> people are putting data into memcached that is mission critical.  I 
> would never put that in memcached.  I would push that data to redundant 
> storage so that all copies were updated properly.
> Brian.

More information about the memcached mailing list