Showing results for 
Search instead for 
Did you mean: 
Showing results for 
Search instead for 
Did you mean: 
Showing results for 
Search instead for 
Did you mean: 

Current Server Hardware Recommendations


Current Server Hardware Recommendations

Anyone know where I can find the current Windchill server hardware recommendations? Any recommendations you can share from your own experience?

We only have 3 Pro/E users and about 10-15 other users. Everything will be on one server (database and fileserver).

# of Processors?
64 bit?
Which Windows Server OS?
Other things to consider?

Currently on 9.1 but will want to upgrade to 10.0 sometime this year.



This page on tech support is a great resource for the information:

Hi Greg,
We just upgraded to Windchill 10. The server is a HP ML350. 2 E5645 2.40GHz processors, 80 GB RAM, 2 TB raid 5, Server 2008 R2 Enterprise, 64 bit (that's important because the standard won't recognize all the RAM). Works great.
We have 5 Pro/E users and 5 Solidworks users and 10-15 other users and yes, database and fileserver on this machine.
Hope that helps.

Mike Pinkney
Triple Ring Technologies.

I use blades. See attached. Depending on the number of users or what you do. I usually apply put a rule of 1 foreground/background methodserver per core. More cores per socket is better and faster. If you can put 4GIGs of ram is adequate per core butto 8GIGs per core is really good. Sometimes you need about 6GIGs to get you past a really big query or other large hitters like a Project Update.

Windchill Server:

  • I suggest now a 2X 6-core or 2X 8-core socket for your windchill server

  • with at least 96 GIGs of RAM with Red Hat Linux 6.2 or 5.7

For your Windows 2008 R2 Server CAD worker:

  • you can have also the 2X 6-core or 2X 8-core socket (1 core per CAD worker) ProE only uses 1 core.

  • 48 GIGs to 96 GIGs of RAM (4GIGs or 8 GIGs of RAM

  • 500GIGs HD, the Windows OS 2008 R2 alone with patches will take up a ton of space

  • Having additional background method servers for publishing is good. Having 1 CAD worker per 10 CAD users is the best ratio. So, 6 CAD ProE workers on on single blade for 60 CAD users is optimal. Thus no waiting for publishes. CAD publishes actually hit the foreground methodserver because it behaves like a user with temporary workspaces.

Oracle Database:

  • For your database, I still recommendOracle but you'll only need a 2X 4-core with 48 GIGs of ram.

  • when you createproduction database, have your memory settings to 40% of the total RAM

  • Always have 30% free space on tablespaces and always avoid auto extending. Just enough space to for time to buy more HDs.

All installations to be placed on the NetApp/SAN/EVA/XPS/etc with all installs as aliases so you can just pull out the machine and pull it out and replace with a DNS alias change. Place all symbolic links and /usr into the SAN as well. That way you don't lose Java, Oracle links. Remember to save your profiles and OS parameter changes.

Externalize your vault and start considering using UnionFS, auFS or SquashFS so that you don't have to copy entire vaults or even entire Windchill installs. Your development server can just be server specific property filesand a separate installation of Cognos (or tailor your certificates to the server). You can place patches and service packs on a different layer without the need to make a complete backup. The intitial install will be placed on the NFS applicance. Just make sure you delete all the logs, and cache as a foundation base. Then the cache and logs will be server specific. This can also be applied to clusters where now you don't have to worry about class files being not sync'ed between servers. There is only one copy on the NFS appliance base and each slave is just property file layers which is server specific. This will save you tons of money in removing duplication of files accross systems which adds $$$ of storage requirements.

My system is running super fast with over 2000 publishes a day. Don't forget to put a test and development systems.

If you use a blade system, all your 16 possible blades can be connected together witha 10GIGE backbone switch which allows all servers to communicate at 10GIGe. You can then have your users connect to the core at 1GIGE.

Having fun,