Convince my boss to use DRS, etc...

[Jacob_A]

Hi,

I'm a Systems Engineer and recently started a new job. When I was finally given access to the entire vmware environment I found some optimizations that could be done.

The main cluster is a 9 host cluster, fairly beefy machines (from 150gb to 384gb of ram, and varies from 1 or 2 physical procs) and each either a Gen7 or Gen8 HP Proliant. These are all connected to a NetApp SAN via fiber.

The first thing I found is they are not using DRS in this cluster, only HA. Yet below is their cluster resource utilization; not horrible, but could be better balanced using DRS.

Next I found that a majority of their VMs still have virtual floppy drives and some have virtual USB controllers, a total waste of resources.

There are also some sloppy things I found, like one host still has IPv6 enabled - we're only using IPv4, so no need for v6 to be enabled. Some of the NICs on the hosts are set to auto-negotiate, which I've always been told not to use and to set an explicit speed. Almost all of the VMs are using the E1000 version of virtual NICs. The power settings on every host is set to 'Balanced' and could be set to High Performance (I've never been in a place where a vm host wasn't set High Performance). And pretty much all of the hosts haven't been updated in about 6-8 months.

Lastly when I was reviewing the licensing I found this main cluster has 6 hosts only with a Standard ESXi 5.5 license applied and 3 have Enterprise, yet there are sites across the world that have a single VM Host and its using an Enterprise license. We only have a limited amount of Enterprise licenses available and most of them are taken up by a single stand-alone host, but could be better utilized in our main cluster at the IT HQ.

I've brought up these items to my boss saying how the environment would run better, but he thinks since things are running fine now, no need to change things. He says better to utilize hours working on something that's going to save us money, not 'make things run a tiny bit better'.

Now I am not swamped with work by any means and changing these items certainly wouldn't take many hours to accomplish.

I dont know how to make a good case for my arguement to convince him to buy into my plan.

Can anyone give me some pointers?

[antonmajor]

The name of the game is this: Find ways to dramatically reduce cost and build a case to present your DRS for further cost savings.

First, you need to show your boss that you have saved him/her XYZ dollars in various places where possible. I think presenting a case of purchasing licensing is then an easier sell.

For the sell, you would need to put it into practical terms from a business perspective.

HYPOTHETICAL:

The IT department starts to receive complaints that certain applications are running slow. You investigate and reveal that there are issues with over utilizing resources on a specific host and cause you to manually intervene. this takes you away from the "Awesome Time Saving, Money Saving Project" you are working on to make the team/company profit further.

You are not fibbing in any way... You should be always looking to better the IT and company you work for. So you should be busy and not be bothered by such trivial items as manually vmotioning a vm that could have done it by itself.

Good luck!

[Alistar]

Oh wow, guess whoever was there before you who established this "mess" hasn't heard of hardware standardization, has he? Unfortunately when your boss doesn't share the "let's optimize things" approach unfortunately (and of course) he will believe "let's save money" approach.

First things first: Keep in mind that DRS' ability to balance load between hosts will be severely impacted if you run it in an imbalanced cluster like this. You need to have the most identical HW configuration you can - do you have any opportunity to schedule a HW maintenance window and move the RAM sticks (if they are compatible) around to ensure a better memory balance? Also keep in mind you will have to estabilish a baseline for your Enhanced vMotion to the "dumbest" processor.

Your idea with resource optimization of removing virtual floppies and USB controllers is fine, this can be given as a "we save resources to host a few more virtual machines on the hosts" reason. Also, whenever power saving kicks in your ESXi hosts will be impacted, so will be the users - this way you have a "we will ensure that the service provided to the business is provided with the performance it needs" reason. The vmxnet3 NICs give you the best throughput, therefore you can also0 say "we will have more stable/faster/optimal networking on our Virtual Machines" plus at any time you can "get affected by a well-known PSOD while using these NICs VMware KB: VMware ESXi 5.x host experiences a purple diagnostic screen mentioning E1000PollRxRing an...".

Everything is running unpatched? There are security issues that need addressing - does your boss want an insider to seize a control of an ESXi host and cause mayhem? As earlier stated by antonmajor, DRS will save on Operating Expenses again by you not having to migrate VMs manually when users/application owners complain about slow performance when a host is overloaded.

And in the end, you just re-balance the licenses, applying enterprise license where it is due and standard license to standalone hosts on the remote sites.

I hope your boss will nudge a bit eventually and let you optimize things by saving operating expenses / man hours that would be caused by decreased performance by automation and optimization in your environment.

[vThinkBeyondVM]

This is what I could list out.

1. Using DRS : You can configure DRS affinity rules for below use cases :

                >. There are NTP servers Say NTP-1 & NTP-2. You can configure DRS affinity rule to keep these VMs away from each other. If one NTP server fails or host fails. Other is available.

               >. Keeping 2 VMs away from each other which are very resource intensive.

               >. 2 VMs are communicating each other frequently, it is better to always keep these VMs on same host. This will help to reduce latency/bandwidth required.

               >. Oracle licensing is great use case of Vm-host affinity rule

2. Effective use of cluster resources . Initial placement + run time placement + maintenance windows : Reduces lot of operating expenses.

3. DPM saves lot of power cost when hosts are not in use without compromising the performance and availability.

4. There is interop with HA + DRS + DPM + EVC  support available.

5. Automation/optimization is the key to success in any organization: