ERP - Virtualize infrastrucure

[WesterWare]

Anyone have experience virtualizing Epicor 9/OpenEdge db?

We are considering Windows Server guests on VMWare with a EMC NX4 iSCSI SAN.

While we are running Epicor 9 in a test environment, we cannot simulate

a real-world load (75 users). Considering the Epicor hardware sizing

recommendations for Epicor 9 (40-150 users), we are worried that

performance may suffer in a VM infrastructure. Anyone?

[Terminal111]

It is a somewhat hard question to answer because you do not give us enough information. Like what all your hardware is, what connectivity you have to the NAS, how many servers you are running on it, etc. I don't have experience with Epicor's product but my advice would be not to try it on a NAS not to mention 1 GbE iSCSI attached NAS. I am virtually certain your performance would suffer. In my opinion, iSCSI is suitable for only low utilization setups even in a SAN. I am doing a similar move with another ERP product at the moment and using Fibre Channel on a SAN it is the only way to go in my opinion.

We looked at the AX4 Clariion and I am instead going with HP equipment and a P2000 with dual connected 8GB Fibre Channel. The throughput of that compared to 1GbE iSCSI SAN is double or more - on the same equipment. If you are using a NAS it will be worse. Reads it absolutely blows iSCSI away. In a 60/40 split or reads to writes it is still double in a RAID 5 or 6. I have about 60 active connections and I believe we do far more reading to writing - Maybe 75/25.

[WesterWare]

Thanks Terminal111 for your comments...

what all your hardware is, what connectivity you have to the NAS,

We would not attempt to run off NAS. As stated, we may be using a EMC NX4 SAN with quad 1GB iSCSI and 22 15K SAS drives (primary storage RAID 10 - block not file based).

I cannot disagree with your stance on iSCSI but if it is true that iSCSI would not be up to the task then that would mean all of the EqualLogic units out there must be running low utilization requirements as that is the only connectivity choice with those units.

Servers under consideration are Dell R710's with dual x5680 (6 cores x

3.33GHz), 48GB RAM, 8 - 1GB NICs w/TOE and iSCSI offload.

[VirtualWatts]

You asked for experience and we don't have that, yet -

Right now we are virtualizing several ERPs -

We are using R610 dual hexa-core 48GB servers with vSphere 4 embed, 4 NICs, iSCSI SAN on MD3220i with 24 disks.

I have 2 NICs/4 ports dedicated to the iSCSI (with ISOE activated), the SAN has 2 controllers/8 ports and 6gb/s backplane.

Although the NICs have TOE and ISOE my understanding is that VMware today does not recognize and use TOE but will use the ISOE.

We couldn't get the EMC into the country fast enough and support was iffy - so it is iSCSI for now.

From what I have gleaned from Dell, VMware and this forum I am more likely to have performance problems within VMware from an incorrectly configured or limited networking arrangement than I am with the actual SAN.

I expect I'll have as many questions as answers but we should be running full capacity tests in about 4 weeks.

I am guessing the only reason you went with the R710 instead of the R610 is to get the extra NICs?

Rick

[WesterWare]

Thanks for the input Rick,

R710 is a bigger number than R610! Actually, we are getting a very good deal on the R710 and like the extra square footage.

Would be interested in your testing evaluation... we won't be ready for testing for a couple of months.

"What weighs more: one pound of bricks or one pound of feathers? Which is

faster: 2 Gb FC or 1 Gb Ethernet?" My reseach is indicating similar performance between a "properly" configured iSCSI SAN and an FC one. An FC SAN will have more bandwidth but bandwidth does not mean better performance. We like the option of switching to FC if we get to a point of saturating our 2-4GB iSCSI pipe.

[VirtualWatts]

So you'll appreciate the DC guys are having a good laugh - when I spec'd the servers I included internal 15k drives.

Habit I guess, totally forgot that nothing would actually be on those drives with the vSphere embed.

I told them they should be glad I didn't order R710's as we'd have a whole bunch of extra drives on our hands.

[bstollfus]

We are currently going through Epicor 9.05 rollout.  We have a DL380p Gen8 with dual E5-2670, and 256GB RAM, with a 765GB FusionIO card.  We have both the SQL2012 server, and the APP server on Fusion, and Epicor is still slow.  We thought it was our SAN HP P4500, but after moving APP server to Fusion with SQL on the same host, we have determined the poor performance is mostly likely within Epicor's code/configuration.

Just to add, we were hitting 950MB/sec off the fusionio card using SQLIO and the tests Epicor gave us, with 0ms latency during the entire test.