TeamSite 6.7.1 on a Sun SPARC Enterprise T2000 server?

Has anyone installed TeamSite 6.7.1 on a Sun SPARC Enterprise T2000 server? If you have, would you be willing to provide some feedback on the performance and any issues you have encountered?

If you could specify the number of cores, CPU speed, memory size and disk type (SATA or SAS), that information would be very useful too.

We are considering the 'Try and Buy' offer from Sun on the T2000, but as always, real user feedback is most valuable.

Thanks,

David

Find more posts tagged with

Comments

jed

Has anyone installed TeamSite 6.7.1 on a Sun SPARC Enterprise T2000 server? If you have, would you be willing to provide some feedback on the performance and any issues you have encountered?

If you could specify the number of cores, CPU speed, memory size and disk type (SATA or SAS), that information would be very useful too.

We are considering the 'Try and Buy' offer from Sun on the T2000, but as always, real user feedback is most valuable.

The t2000 is a great box for specific tasks such as web serving. We run t2000's for all of the sun.com backend appservers. Soon, we will use t1000's for our front-end webservers. In general, these boxes are great at scaling up for applications that like to thread.

These boxes do have some weaknesses. For example, they suck at:
o Floating point math
o Single threaded applications (low overall mhz rating)

We ran some scalability tests with a t2000 this past December with TeamSite 6.7.1. Some tests ran very well. For example, our publication process did a pretty good job of scaling up. This consists of a bunch of external tasks that set metadata, regen templates, submit, create filelists and deploy.

On the other hand, core server operations lagged and did not scale When testing core TeamSite operations such as reads/writes/submits/etc, the time to complete the various operations increased significantly as parallel operations increased. It appears many core operations are single threaded. (The number of mutex locks as alarming.) Given the low single thread speed of the t2000, the box may not be the best for TeamSite.

Is a t2000 good for you? It all depends on how you use your TeamSite server. The great thing about the try and buy program is that you can run your own tests without infrastructure costs. If you don't like it, return it Smiley Happy

We are upgrading our production server soon. We decided a t2000 would not be good for our usage patters. We are going with a M5000 due to the much faster processor speed. The box is ****-expensive, but I tend to get good prices on hardware Smiley Happy

http://www.sun.com/servers/midrange/m5000/

For your needs, a v490 or m4000 could work out better if you are sticking with sparc.

In my opinion, the "best" boxes for TeamSite would be the x4600:
http://www.sun.com/servers/x64/x4600/

This has fast CPU's to quickly process the single threaded operations and lots of CPU threads to run all the misc. workflow tasks that generally run on a TeamSite server. Unfortunately for me, Interwoven does not support x86 Solaris and Linux is not an option for me.

The full specs of our test box were:
o 1.2 mhz/8 core
o Fiber channel disk storage
o 32 GB RAM

Hope that helps...
~jed

Migrateduser

We ran some scalability tests with a t2000 this past December with TeamSite 6.7.1. Some tests ran very well. For example, our publication process did a pretty good job of scaling up. This consists of a bunch of external tasks that set metadata, regen templates, submit, create filelists and deploy.

On the other hand, core server operations lagged and did not scale When testing core TeamSite operations such as reads/writes/submits/etc, the time to complete the various operations increased significantly as parallel operations increased. It appears many core operations are single threaded. (The number of mutex locks as alarming.) Given the low single thread speed of the t2000, the box may not be the best for TeamSite.

May I ask how did you do these scalability tests? Did you use any commercial load testing software? Thanks.

jed

May I ask how did you do these scalability tests? Did you use any commercial load testing software? Thanks.

I created custom scripts that would run a test and ramp up the number of parallel operations. I only looked at the time to complete the various tests. Example:

Test 1
o Run test with one thread to completion three times in a row

Test 2
o Run test with two threads to completion three times in a row

Test 3
o Run test with three threads to completion three times in a row

and so on...

When running tests with multiple threads, I always used different branches. Example:
thread 1: /default/main/tests/branch1/WORKAREA/foo
thread 2: /default/main/tests/branch2/WORKAREA/foo
thread 3: /default/main/tests/branch3/WORKAREA/foo

Samples of tests included:
o Running our publish workflow
o Running "core" server operations including file copies, listings, deletes, creates, and submits

Note: File reads/listings seemed to scale fine. Creates/deletes/copies got very slow as the number of threads increased.

Ariane

Thanks Jed,

This is exactly the kind of information I was looking for. I appreciate your analysis regarding the single thread performance.

We'll take a look at the x4600 and the M5000. Price may be a problem as is often the case. Our requirement is for 4 servers.

Have you ever tried virtualization on something like an M5000? It was suggested to us that might be an alternate way of implementing multiple server environments.

Cheers,

David

nipper

Have you ever tried virtualization on something like an M5000? It was suggested to us that might be an alternate way of implementing multiple server environments.

Cheers,

David

I installed TS 5.5.2, MT and OD on a virtual slice of an older (circia 2003) mongo Sun server. I did not try to install multiple instances but a single instance worked fine. MT was a pain to get going, but eventually I got it correct.

Andy

jed

This is exactly the kind of information I was looking for. I appreciate your analysis regarding the single thread performance.

We'll take a look at the x4600 and the M5000. Price may be a problem as is often the case. Our requirement is for 4 servers.

Have you ever tried virtualization on something like an M5000? It was suggested to us that might be an alternate way of implementing multiple server environments.

We don't have the M5000 yet, so I cannot speak for that machine. Historically, we have not done any virtualization with a TeamSite server. If the M5000 supports hardware based virtualization, that should work--in theory.

If you do not need a ton of horsepower, you should also look at the m4000. They are cheaper and more importantly, smaller. (6RU vs 10RU)