An alternative to Tugela cache

[Alberto Bertogli]

On Sun, Sep 02, 2007 at 11:37:33AM -0400, Perrin Harkins wrote:
> On 9/1/07, Alberto Bertogli <albertito at gmail.com> wrote:
> > It's non blocking and can use bdb or qdbm as backends.
> 
> The single database thread, while much easier to design, seems certain
> to be a bottleneck.  BerkeleyDB definitely supports the locking
> capabilities to handle multi-threaded access.  Any thoughts about what
> it would take to change that?

Well, having multiple worker threads wouldn't be too hard to do, but it
would need very careful design not to break the ordering assumptions.
I'm not sure about how much performance could be gained with this
approach, but I'll look into it. If you have any BDB performance numbers
comparing single vs. multiple threads, please let me know.


> Also, have you done any speed comparisons with memcached, direct
> BerkeleyDB, or MySQL?

I did quick tests against memcached using the TCP protocol over
localhost some time ago and they were about the same speed, but it
wasn't a very detailed test. I haven't done any tests comparing it to
direct BDB or MySQL yet, but it's a good idea, so I'll try to write some
this week.

Thanks,
		Alberto