We are planning to upgrade our Baan server with Xeon technology & Windows 2000 Server OS.
I heard we might end up with problems if hyper-threading is enabled in the server.

Is anybody using hyper-threading with Baan IVc4 & Windows 2000 Servers?

Any feedback will be much appreciated

Thanks

Hi,

We have BAAN IV c4 on IBM x235 whit hyper-threading and we dont have any problem's

For hyperthreading no special coding need to be done. If an application runs on a multi CPU system (as Baan is doing) no issues are to be expected on a hyperthreading system.

Have no ideas where you heard these rumors, but I can't image any reason for problems in this area.

Kind regards,
Dick

This article has a good explanation on how this works: http://www.2cpu.com/articles/ht_explored/index.html

There is a danger that the OS will incorrectly allocate resource on a multi-threaded application, leaving 1 CPU empty and another 100% (in a dual CPU box). I have seen benchmarks that suggest performance benefits from turning HT off.

It may be worth testing...

I had the opportunity to do a benchmark at Dell earlier this year as well as have several client running with HT. It has a very positive effect on performance and overcomes some of the earlier restrictions on minimum numbers of processors given a batch load on the system. I have seen decent performance on two processor systems with a light batch load -- definitely not the case without HT.

SO yes -- HT good.

I have seen benchmarks that suggest performance benefits from turning HT off.

It may be worth testing...

I agree with JamesV. I wonder if ythe benchmarks you have seen are Baan/SSA benchmarks or other benchamrks as on 2cpu.com

Kind regards,
Dick

You need two thinks to take an advantage of HT:

1.: Your OS must support HT ( Linux 2.6.xx or late 2.4.xx, W2k3 )
2.: Your application should support HT too.

So neither W2k nor the Baan binaries support it!
I `d made some tests with a 4-way server with W2k and Corelli. I win 20%-30% performance switching HT off.

:rolleyes:

gonzmaus,

I can image that your results are correct. You probably ran a few (up to the number of CPU's you have) batch processes in parallel. We use the rule of thumb that a HT CPU is in fact 2 CPU's of 60%. When making full advantage of it (Full Load with OLTP), you have in total 120% of the CPU power compared to switching HT off. This explain why you measured no improvement when HT was on. The amount of parallel processes was too low.

So, when only having a few processes in parallel, HT can decrease your performance, when having more (like OLTP), your system will gain from HT.

Hope this helps,
Dick

BTW: As you see above, your application doesn't need to support HT to benifit from it. And believe me, we have done tons of benchmarks here.