Measuring VMWare performance - (more than just the CPU cycles)

[cliffskygod]

VMWare performance –

Cluster:

Hosts: 2 proc, 4 core, 2.659 Zeon X5550, 96GB Ram  :  5 host Cluster (Total 106360 Mhz)

Our VM cluster stays around 50% (45-65%) day/nights/weekends – why?

No application servers except –

(2) MSSQL servers (10 users access, low bandwidth)

(6) Network Monitoring/taps –

(3) AD servers – for 100 users

(2) Print Servers – for 100 users

(20) Light Win-7 desktops (used for outlook only)

(20) Linux desktops

(8) VM’s used for Avamar backups/tape out

(40) Server 2008/12  - not doing anything if you check local taskman/perfmon

vCPU/Core = 6.89 to 1 = Is this too high?

Mem: 455GB Granted, 347GB consumed, 46GB active

Why (how to record/measure) a client’s load on a host? What is running within the client that is requesting the resource? i.e. How can I tell that cycles used on my hosts were actually used/required by the client?

How/where are the CPU/Disk/Network tracked for backup using the VADP API calls?

Efficiency/effectiveness of Hyper-threading? How can you tell if clients are taking advantage of HT?

Looking for the best reason why our VM environment uses "X" amount of resources. If the answer is: "The clients use what resources they need" - that is fine, but documentation would be nice.

Thanks, Cliff

[MattG]

Have you looked into vROPs?   If not,  at minimum you could install it in trial mode for 60 days.   It is a simple install and will give you this data and more.   Combine it with Log Insight (included with vCenter 6 license) key and you will know what is going on.

-MattG

[cliffskygod]

Matt -

Is there a quick setup for vROPs/Log? I installed, but looks a bit intense.

Once configured, how long to start having enough data to make sense? 14, 30, 45 days?

Thanks again,

-cliff

[MattG · Accepted solution]

vROPs is a single virtual appliance.   It is a simple install:

• Deploy OVF
• Connect via browser
• Choose Simple method and select trial license
• Login as Admin
• Go to Administration->Solutions->VMware vSphere->vCenter Adapter
• Configure vCenter adapter (vCenter FQDN and credentials for vROPs to pull data from vROPs)

Thats it.  Within 5 minutes it will start pulling all vCenter metrics.

Data will be useful as soon as it has enough to meet your analysis need.  vROPs does dynamic thresholds which take longer to start generating,  but you can look at the metrics in the Troubleshooting ASAP.

-MattG

[cliffskygod]

Thanks Matt - collecting data now.

Any additional steps for log realize? I installed, but nothing has asked for configuration besides initial install (name/ip etc.).

[MattG]

In Log Insight go to Administration (top right corner) -> Integration -> vRealize Operations

Enter the vCenter host name and password.

To pass data from LI to vROPs,  create an LI query and Create Alert from Query, Send to vRops.

-MattG

[cliffskygod]

Thanks again Matt - lets see what data it collects and if it is useful.

How probable is it that the resources are being used by unattended applications like: windows updates, anti-virus, backup are causing CPU usage over 50% during off hours when no one is connected?

Any others I am leaving out?

We do not run any "application servers", but do have a number of windows desktop and server clients (75) - no VDI, another 35 linux.

No issues with the Linux systems running amuck on their own.

Thanks, Cliff

[MattG]

Processes usually run on a schedule.   These schedules most likely run on the hour (maybe 1/2 hour).   If you see spikes that begin on the hour on a regular basis there is a good chance that its a process.    You can gather further insight into these processes by looking at other performance metrics like network usage.

-MattG