AnsweredAssumed Answered

EPDM Slowdown When File has Many Versions

Question asked by Jim Sculley on Jun 16, 2016
Latest reply on Jun 17, 2016 by Charley Saint

Like • Show 0 Likes0
Comment • 3

I've been working on an add-in and during testing, I have some test files that I test over and over (state changes, check in, check out, etc). I have noticed that as the version count increases, certain EPDM operations slow down significantly. For example, clicking the 'Change State' button in the EPDM task pane in SOLIDWORKS would result in a five second delay before listing the available transition (there is only one for this test). This was a file with 300+ versions. As soon as I performed a Copy Tree to make a new file, the response time was instantaneous.

Is this a known problem? I imagine that many EPDM users have files with lots of versions, especially if they have hardware libraries and such. I don't know if the impact would be multiplied if the file is part of an assembly.

1 person has this question

Outcomes

Prasad Bhonsule Jun 16, 2016 5:31 PM

Hi Jim, looking through the Knowledge Base, it looks like this was a known issue in EPDM 2011 and 2012, but was marked as fixed in 2013 - see SPR 629573

If you're still experiencing this behaviour, I would suggest raising a case with your VAR, I would imagine they have some more information or can raise it with SW.

Thanks

Prasad

Actions

Tim Webb Jun 17, 2016 2:06 AM

Jim,

Is your SQL DB server virtual? If so, I have one client with daily versions and transitions have created hundreds of their core files with 250+ versions and their vault was slow!

Are you experiencing head blocks when that file's contains tab is activated or if a transition dialog has to build the reference tree? If you were to trace it, the offending SP is probably dbo.XRef_GetTree. Check out SPR#792546, 788790, 802067, 901322, and 758929 We found this SP to be the root symptom while searching for the root cause. At one point we tried a un-encrypted version of that SP provided by SW and it helped but not significantly.

We researched for about a month and finally talked with my VM guru guy who asked the client a few questions as he had encountered this before.

"How many hypervisors are on the cluster? Is the SQL DB HV in a cluster of mutliple VMs? If so, the DB may be ready to run but can't get processor time so your VM guy can set up a configuration in the cluster to migrate the other workloads off that host for 24 hours to see if there is sufficient improvement before committing those changes. It's a delicate balancing act but should work."

Basically SQL DB is in a ready state ready to run but it is always fighting for processor time on a cluster where other servers are getting all the processor time and when there is a huge query hitting the DB by way of the XRef_GetTree.

LOL I'm totally in over my head here. Charley Saint can probably shed some light on what I'm trying to articulate. My IT resource guy showed my client how to fix it and they ran a test for a 24 hour period and it worked but I don't know how it is doing today.

Hope this helps,

Tim CEPA

http://www.equivaq.com

Actions

Charley Saint @ Tim Webb on Jun 17, 2016 8:05 PM

I've fought with XRef_GetTree more than I care to admit, there are a lot of variables that get blended into that SP and nearly everything in EPDM relies on it. Processor latency typically affects smaller jobs that run a lot more than longer running ones. Think of it this way, if you have to wait 15ms each time for a task that runs in 5ms then it'll run 300% slower, but 15ms on a 150ms job will only run 10% slower. I often run test db's with some unencrypted SP's because it unlocks a bunch of performance monitoring tools, specifically the Database Engine Tuning Advisor.

Actions

3 Replies

Prasad Bhonsule Jun 16, 2016 5:31 PM
Mark Correct
Correct Answer
Hi Jim, looking through the Knowledge Base, it looks like this was a known issue in EPDM 2011 and 2012, but was marked as fixed in 2013 - see SPR 629573

If you're still experiencing this behaviour, I would suggest raising a case with your VAR, I would imagine they have some more information or can raise it with SW.

Thanks

Prasad
Like • Show 0 Likes0
Actions
Tim Webb Jun 17, 2016 2:06 AM
Mark Correct
Correct Answer
Jim,
Is your SQL DB server virtual? If so, I have one client with daily versions and transitions have created hundreds of their core files with 250+ versions and their vault was slow!

Are you experiencing head blocks when that file's contains tab is activated or if a transition dialog has to build the reference tree? If you were to trace it, the offending SP is probably dbo.XRef_GetTree. Check out SPR#792546, 788790, 802067, 901322, and 758929 We found this SP to be the root symptom while searching for the root cause. At one point we tried a un-encrypted version of that SP provided by SW and it helped but not significantly.

We researched for about a month and finally talked with my VM guru guy who asked the client a few questions as he had encountered this before.
"How many hypervisors are on the cluster? Is the SQL DB HV in a cluster of mutliple VMs? If so, the DB may be ready to run but can't get processor time so your VM guy can set up a configuration in the cluster to migrate the other workloads off that host for 24 hours to see if there is sufficient improvement before committing those changes. It's a delicate balancing act but should work."

Basically SQL DB is in a ready state ready to run but it is always fighting for processor time on a cluster where other servers are getting all the processor time and when there is a huge query hitting the DB by way of the XRef_GetTree.

LOL I'm totally in over my head here. Charley Saint can probably shed some light on what I'm trying to articulate. My IT resource guy showed my client how to fix it and they ran a test for a 24 hour period and it worked but I don't know how it is doing today.

Hope this helps,
Tim CEPA
http://www.equivaq.com
Like • Show 0 Likes0
Actions
- Charley Saint @ Tim Webb on Jun 17, 2016 8:05 PM
  Mark Correct
  Correct Answer
  I've fought with XRef_GetTree more than I care to admit, there are a lot of variables that get blended into that SP and nearly everything in EPDM relies on it. Processor latency typically affects smaller jobs that run a lot more than longer running ones. Think of it this way, if you have to wait 15ms each time for a task that runs in 5ms then it'll run 300% slower, but 15ms on a 150ms job will only run 10% slower. I often run test db's with some unencrypted SP's because it unlocks a bunch of performance monitoring tools, specifically the Database Engine Tuning Advisor.
  Like • Show 1 Like1
  Actions

EPDM Slowdown When File has Many Versions

Outcomes

Related Content

Recommended Content