Best setup for upgraded system, splitting up data-logs-index

[scbrigad]

I tried to import my database data directory on to another computer again with no luck.. Has anyone done this when porting their database to a different computer or is the database tied to the current system configuration? I.E. Is my database recoverable with the new system config or on a totally different system?

[kraada]

So long as the other database was from an 8.3 version of PostgreSQL it should function regardless of the hardware it was installed upon. If it was an 8.2 version though you need to actually make a database backup and restore it into 8.3 as 8.3.x cannot read 8.2.x databases.

[yournamehere]

I'd be happy to run any benchmarks you'd like to see, Miami, are there any particular settings you'd like me to use with pgbench?

kraada -

with your expertise (or maybe ask josh) can you recommend which pgbench params would be a ballpark emulation of pt both 1) importing hh and 2) hud/tracker reading from dbs? specifically I'm interested in "-c N", "-C" and "-M" (http://developer.postgresql.org/pgdocs/ ... bench.html).

I v/a a 150M test db before each run, am using -s 10 and -t 10000 but if 1) pt accesses as a single client ("-c 1") and/or hud/tracker precludes using "-C" then I see fairly wildly different results which obv greatly impacts some of the (software) tuning suggestions.

so iow if I can find out how to roughly model the actual way tracker/hud accesses/uses the db then I can correctly set pgsql cfg.

it would also be nice to know to what extent, if any, db size is a factor (like maybe the test db needs to be hundreds of megs instead of one hundred for example).

thanks

[APerfect10]

I'm not sure how reliable pgbench results will be trying to simulate PT3 unless if you write your own test scripts to emulate PT3. Anyways...

1. Import
   1. -c (1 connection)
   2. -C (uses same/persistent connection)
   3. -M simple
   4. -t (100 hands per transaction; which depends on number of players at the table and whether its a tourney or cash)
2. Tracker/HUD
   1. -c (2 connections per database)
   2. -C (uses same/persistent connetions)
   3. -M simple

Best regards,

Derek

[scbrigad]

ok, running pg_resetxlog fixed the crashing problem I would have... so apparently using 'junction.exe' to move my log directories to another drive somehow corrupted the data even after removing the dynamic links. I suggest everyone be careful and do backups when creating these logs.

Its a hard thing to compare results without a real benchmark as being discussed... but I am getting 400-500 hands/s imports (750+ h/s on fresh database) running on my setup indicated earlier with the exception of only having 2gb 800MHz DDR2 at the moment and overclocking the processor to 3.8 GHz 1600 FSB. No tweaking of postgresql.conf with my current install but I'll play with it if memory still becomes a bottleneck after I install the extra memory. XP Performance monitor does seem to be showing a bottleneck with pages/s loading. This is running the Beta 23 version of PT.

-Scott

[MiamiVIce80]

@scbrigad

i tried junction.exe and linkd.exe on a XP system and it didnt worked. Looks like the mklink.exe from Vista is able to create directory junctions. Not sure if there is a tool for XP that offers that as well. At least junction.exe and linkd.exe didnt look like they do. Maybe someone who was able to get it running on XP could help.

Maybe we could develop a benchmark for Postgres, but that means we could need some details from the developers and i guess they will not offer user there secrets

but a pgbench setup just for testing would be nice. I´m not used to pgbench. Maybe someone could write a small script and we can go testing an compare our results.

(no idea what scale_factor or fillfactor i should use during initialization of the testdatabase, benchmark: pgbench -c 1 -C -M -t 100 (fromAPerfect10))

[Eric Poker]

Tell me if i'm wrong but few month agos the drivers in XP/Vista was not really good to "transfer" small Files to a SSD. In fact, 7200rpm HD was 2x faster in "writing" then a SSD

[APerfect10]

Tell me if i'm wrong but few month agos the drivers in XP/Vista was not really good to "transfer" small Files to a SSD. In fact, 7200rpm HD was 2x faster in "writing" then a SSD

That is correct that XP/Vista are not really optimized to take full advantage of the SSD's speed; however, I do not think a 7200RPM drives write performance is 2x greater...although I guess it depends on the quality of the SSD drive. There is a wide range of SSD's available and their performance varies greatly (typically the more expensive are much faster).

Best regards,

Derek

[scbrigad]

Hmm.... you may be right about the XP drivers, I might be forced to switch to linux or Windows Server after all if postgresql.conf tweaking dosent help. Housekeeping is still really slow, as I am not seeing reads get much over 20 MB/s on database cache updates and writes not much over 4 MB/s even now that I have maxed my memory at the XP limit 3.2GB and runinng an enthusiast level quad-core processor.

Certainly the X25-M performs well in benchmarks like h2benchw 3.6 (consistent 200MB/s read, 75 MB/s write) and crazy database IOMeter 2003.05.10 IOPS in the 3500 range running Windows Server 2003. (http://www.tomshardware.com/reviews/Int ... ,2012.html)

Is anything giong to change in Windows 7 or should I start porting over to a more server-friendly OS now? Can I run Windows Server and not lose too much loading multiple Virtual Machines running TinyXP? I am currently running PT3 auto-imports directly from those VMs so maybe I would need to run it straight from Windows Server...

@MiamiVice80 What are you running right now? Based on this quote

"
I´m not sure what i will do but the obly way to get real performance is to run postgres on a Linux-Server and use an cheap office PC as terminal running Windows the get access to your poker site.

Or u run postgress on Linux and use a VM to run Windows and play. Some people already announced that they run that kind of system.
Maybe someone using Linux+Windows could post some details how fast his setup is by running pgbench or import 1mio hands into a big database.
"

I am guessing you think Windows Server can't do anything near the job of running a Linux server? I haven't seen anyone chiming in on their pgbench results of Linux+Windows, which thread are you referring to? Are there any future plans to support PT3 running directly in Linux or Windows Server to avoid having to use a VM?
I

[scbrigad]

Well it sounds like this won't compare to a linux-server solution, but Windows 7 with some SSD specific speed tweaks will certainly be faster for SSD's than XP or Vista. If Microsoft does more to optimize them before the official release or future service packs, it might be a workable solution to be able to operate postgres and PT3/poker apps in the native OS at decent speeds.

http://www.cdfreaks.com/news/15206-Wind ... -HDDs.html
http://www.computerworld.com/action/art ... geNumber=2
http://mydellmini.com/forum/windows-7-u ... t2540.html