Subject | Re: [firebird-support] Re: Firebird/2.5 and Win/2008R2 |
---|---|
Author | Thomas Steinmaurer |
Post date | 2012-01-28T17:57:37Z |
Hello Aage,
You also could be interested in the following 2.5 architecure comparison
sheet:
http://www.firebirdsql.org/file/fb25_architecture_comparison.pdf
Firebird internal page cache. I haven't done any tests on that, but
AFAIK some others (Paul Reeves?) did. AFAIR also in one of Dmitry's
presentation they compared the efficiency of the Firebird internal page
cache vs. file system cachen and when both are involved, but I might be
wrong here.
With regards,
Thomas Steinmaurer (^TS^)
Firebird Technology Evangelist
http://www.upscene.com/
http://www.firebirdsql.org/en/firebird-foundation/
> Alexandre responded:The same as for CS, so a local cache per connection.
> <<
> My experience is with OLTP use, once you said that there is few
> insert/update/delete perhaps the shared cache of SS could be a good
> benefit, but I think that the OS cache could do the same job...
> A natural choice for SMP machine is CS or SC...
> I would choose SC or CS instead of SS even if the most use is select's
> >>
>
> Alexey responded:
> <<
> Use Classic or SuperClassic to benefit from multi-CPU environment (if
> one user will run bad query which consumes 100% CPU, others will work
> without any problem).
> Since RAM is> database size, the database will be stored completely in
> file cache (after some time), it means better performance, but stability
> becomes an issue - ensure you make reliable backups and, of course, use UPS.
> >>
>
> Thank you for comments and advice.
>
> Are there any peculiarities with the setup of SuperClassic? Cache
> size for CS is very different than cache size for SS - what are the
> rules for SuperClassic?
> The main database (in SuperCerver mode) currently uses a page size ofSee above.
> 8KB, and 8K page buffers - totalling 64MB. This (64MB) isn't much,
> but it is according to the old recommendations of "don't use more
> than about 10000 pages". Does a cache size of 1GB (with 64bit
> SuperClassic) seem reasonable?
You also could be interested in the following 2.5 architecure comparison
sheet:
http://www.firebirdsql.org/file/fb25_architecture_comparison.pdf
> What are experiences with the configuration setting "FileSystemCacheThreshold"?It controls the relationship between the OS file system cache and the
Firebird internal page cache. I haven't done any tests on that, but
AFAIK some others (Paul Reeves?) did. AFAIR also in one of Dmitry's
presentation they compared the efficiency of the Firebird internal page
cache vs. file system cachen and when both are involved, but I might be
wrong here.
> The ReleaseNotes mentions its relationship to the value of--
> "MaxFileSystemCache", but it also says that "MaxFileSystemCache" is
> no longer a valid parameter (for 2.5)! Maybe this needs to be
> cleaned up in the ReleaseNotes?
With regards,
Thomas Steinmaurer (^TS^)
Firebird Technology Evangelist
http://www.upscene.com/
http://www.firebirdsql.org/en/firebird-foundation/