Hi,

we had the same problem ( 7 DMX, but without Vmax). Each DMX have > 30000 Devices (this is an part of this problem). Take a look to the Release notes of SMC. The limit of volumes are 80000 on max 10 DMX.

Next big point in our environment was the GNS-demon. We use devicegroups (> 5000). After deactivating the gns ist fast (grr, better... not so slowly:-)...

regards

Michael

P.S: It's recommended (EMC PS) each DMX/VMAX have it own SMC-Server...

SMC server per array, how robust !!!

Hi,

just another Tip:

the SMC Version 7.0 is faster then 7.1 ! After an Upgrade next step in our environment was an downgrade :-(

Regards

Michael

We have several installs of SMC in our environment and it runs just fine.  We have up to 7 arrays on one instance with 1 V-Max, several DMX3/4's the response time is quite good.  Keep in mind that on a V-Max array if you put the service processor on your network you can have SMC running on it; if you are just doing work on the V-max you can do it all from there.

Ok i´ve fount the Problem.... :-)

Our Unix Guy´s changed the Tomcat Config because of Security Reasons. They deactivated the HTTP Port and Activate the ajp 1.3 Port in order to communicate with an Apache Webserver on the same System. The Apache should be responsable for the HTTPS Requests and route them trough to Tomcat.

I don´t unterstand in detail why they do that, but there is a problem with the Self Signed SSL Certificate in our Environment if i unterstand it correct.

Basicly this will work but the Application shows the Performance Problems i spoke about. After change back to Standard Configuration everything works fine.

Next week we will speak to an EMC Software Guy in Order to figure out what the Problem ist. Maybe someone did things like this before, if yes please let me know.

Hi Martin:

I'm interested what the results of your follow will be.  Keep us informed, please! 

Mike

How many GK you assigned from VMAX to SMC server?

Michael Soltau wrote: Hi, we had the same problem ( 7 DMX, but without Vmax). Each DMX have > 30000 Devices (this is an part of this problem). Take a look to the Release notes of SMC. The limit of volumes are 80000 on max 10 DMX. Next big point in our environment was the GNS-demon. We use devicegroups (> 5000). After deactivating the gns ist fast (grr, better... not so slowly:-)... regards Michael P.S: It's recommended (EMC PS) each DMX/VMAX have it own SMC-Server...

Hi Michael:

I've seen the words 'devices' and 'volumes' tossed about interchangably.

for a Rough guess on devices/volumes/whatever, I do the following for each SID... is this anywhere nearly correct?

symdev -sid XXXX -list -hyper | wc -l

I was amazed that my single SMC server (very slow) handles 9 Symms with 191,000+ devices... but its too slow to be practical, and the 80000 rule, if it applies here, may be the reason behind my slowness.

Proposed Solution to the community:

With the above 'metric' as my count of 'devices/volumes/whatever', I can then spread the load out accordingly.

Present State:

My poor overloaded SMC server runs on a VM guest, with Win2K3 as the OS(2 Gig ram, 1vCPU).  It is proxied to a solutions enabler server which happens to run AIX.  That particular AIX server serves a backup application, and was an easy place to start for this purpose:  it has all these Symms zoned to it, with at least 6 GK's each.

Solution:

From my massive amounts of devices coming back from that set of 9 symms, I have one particular Symm (A DMX3-6 running 5771.111.0 enginuity) that returned 61K+ hypers (!).  My first step would ideally be to spin up another VM, install SMC on it, and give the VM guest 10 Gatekeepers from that symm via RDM's.  I would then stick to the 80000 rule and portion out another VM with RDM's from a selection of other Symms to balance the load appropriately.

Request for Comments:

The thoughts of the community are welcome.

Mike