jmroth
Enthusiast
Enthusiast

vMotion to new hardware (and cluster) results in performance penalty

Jump to solution

Platform 1 is 4 or 6 core X5550 or X5650 @2.6GHz (outside of any cluster) on ESXi 5.1 r1312873

Platform 2 is 4 core E5-2637v2 @3.5GHz (Cluster with SandyBridge EVC Mode) on ESXi 5.5 r2068190

After vmotion from Platform 1 to 2 (database servers) users are reporting performance penalties.

Looking at our Cacti graphs representing CPU usage we see the following. (You can imagine when the vmotion happened.)

vmotion-cpu.jpg

Is this normal? Why is this?

The specific VM I am showing here is a VM in version 7 with 2 vCPU and 18G RAM using SLES Linux (kernel 2.6.27).

Note after complete shutdown and restart, it's ok (tested on other machines affected the same way)

0 Kudos
1 Solution

Accepted Solutions
jmroth
Enthusiast
Enthusiast

Alright, we've now disabled power management in the new hosts' BIOSes...:smileyplain:

View solution in original post

0 Kudos
2 Replies
jmroth
Enthusiast
Enthusiast

Alright, we've now disabled power management in the new hosts' BIOSes...:smileyplain:

0 Kudos
vervoortjurgen
Hot Shot
Hot Shot

hi

my first tought is because you new hosts are 4 cores and not 6 core wich means x5550 or x5560 can have more threads.

but because you mention database server it could be your storage too or path to it. same storage? fiber? iscsi?, is your zoning correct if your on fiber or size for the new cluster?

kind regards Vervoort Jurgen VCP6-DCV, VCP-cloud http://www.vdssystems.be
0 Kudos