> Tell me if I am wrong but it sounds to me like like
> an endless problem....
Agreed. Such it is with caching. After doing some informal
benchmarking with 8.0 under Solaris, I am convinced that our major choke
point is WAL synchronization, at least for applications with a high
commit rate.
We have noticed a substantial improvement in performance with 8.0 vs
7.4.6. All of the update/insert problems seem to have gone away, save
WAL syncing.
I may have to take back what I said about indexes.
Olivier Sirven wrote:
> Le Vendredi 21 Janvier 2005 19:18, Marty Scholes a écrit :
>
>>The indexes can be put on a RAM disk tablespace and that's the end of
>>index problems -- just make sure you have enough memory available. Also
>>make sure that the machine can restart correctly after a crash: the
>>tablespace is dropped and recreated, along with the indexes. This will
>>cause a machine restart to take some time.
>
> Tell me if I am wrong but it sounds to me like like an endless problem....This
> solution may work with small indexes (less than 4GB) but what appends when
> the indexes grow ? You would add more memory to your server ? But there will
> be a moment were you can not add more so what's next ?