Our system is setup for 1000 users including 70 ProE users. I suggest you have at least 24GIGs to 48 GIGs of DDR3 RAM with at least 2X 3GHZ Intel quad core or AMD 2 6-core. I found it great performance when using 40% memory setting when creating your database usingdbca. I also set:
- process to 300
- block sizes to 64K
- open cursors to 2000
- and so on.
Take a look at these settings if you are using ProE. It may conflict with the performance tuning guid for Windchill.
http://www.ptc.com/appserver/wcms/standards/freefull_cskdb.jsp?&im_dbkey=105413&icg_dbkey=900
The only suggestions is I always have the min memory heap to 768 MEG and the max 3968 MEG to the method servers (if you have enough memory). We have 48GIG of DDR3 ram on both the application and database servers. You can always limit the max heap to a smaller amount and add more foreground method servers.
There are now 2X 6-core blade systems which you can get upto 96GIG or 192 GIG of RAM:
http://h71016.www7.hp.com/dstore/MiddleFrame.asp?page=config&ProductLineId=431&FamilyId=2982&BaseId=29501&oi=E9CED&BEID=19701&SBLID=
http://configure.us.dell.com/dellstore/config.aspx?c=us&cs=555&l=en&oc=MLB1817&s=biz
Also, I would always go with Red Hat Linux ES 5.5 or 5.4 (EVA switch issue) x86_64 since it always seems to outperform any other OS.
We also have3 background methodservers, 2 forpublishing queues and 1 default, so that it doesn't bog down the single background method server. I suggest having 1 foreground for every 100 to 150 users. We have 6 foreground. Saves tons of money instead of clustering, The backgrounds are always the single point of failures anyways. If your default background freezes up and restarts and causes workflow issues. This is a lot worsethan your foreground hangs which stops a user session from commiting. I also suggest having 1 CAD worker for every 10 ProE user. We have 6 ProE CAD workers on one machine.
good luck,
Patrick
