Showing results for 
Search instead for 
Did you mean: 

What If Scenarios - how cool is that?

Level 9

One of the cool function of any virtualization management suite is not only to have a nice dashboard showing which VMs performs poorly on which datastore/host, but also to be able to see what will happens if:

  • I create 10 VMs a week - in how many weeks/days I'll need a new SAN or new host(s) (storage capacity planning)
  • I add more VMs -  I increase my workloads by adding 20 new SQL server databases
  • Workload rise - I want to know the storage latency I'll have on my datastore if the overall workload increases 20%.

The prediction possibilities. If you could predict the future in your life you would probably be milionaire right? You know you can't.

However if you want to predict future in your datacenter? Yes, you can! For example by simulating the increase of the workflows you get the whole picture in 2 weeks, 2 months or more. Pretty awesome.

As an IT guy I usually visit existing datacenters so pure green fields are rare. Usually the datacenter, or the client has already some virtualization already in place. In this case I always use some initial workloads which are then taken into consideration to predict the future expansion.

A client of mine asked me recently on how many new hosts he will need to accommodate X new VMs and the grow for Y VMs a day. No longer guessing. The tools are here for such a tasks and if used correctly they give accurate results.

Did you tried digging through one of those tools or you continue guessing?

Level 12

Cool indeed. It is very intersting to be able to create those number of VMs as long as you can be proactive in monitoring them well enough.

I believe you have the right tools for monitoring the number of VMs in your DC.

Level 10

We typically went/go by what the customer "asks" for from us. This is not always the best approach. Even seasoned veterans in the IT industry can get it incorrect when the technology is all new. When the client asks for virtualization they think money saving. This can sometimes be a misconception in my opinion. After licensing, hardware, etc. Sometimes it can be as much, but the long term benefits still outweigh the short term gains.

So in the past we took a very large guess at what was needed, and just assumed the clients all knew what they want. We no longer have to do this thanks to virtualization management. We have an excellent idea of what the current load is, and if they want/need to scale up, or create more VM's, rogue VM's, old snap shots, etc. Excellent tools that I truly believe are completely invaluable for almost anyone.


Level 11

Please let me know the perfect toll for mentioned scenario..

Level 12

+1 in the past, we have gone by what the customer asks for and found out that it was not beneficial on a long term basis.

Level 9

More often then not, I've seen the guessing method.  However, I'm delighted that the mix has shifted more in favor of trying to predict by collecting data and using tools.  Depends on the org, their size and resources available, but they are now seeing that a more data influenced method is more useful than throwing a dart randomly.  Pulling the right data with the right tool is the big thing there though.  Grabbing bad data can sometimes be worse than grabbing none at all.

Level 9

It depends, there is quite a few on the market today including the one from Solarwinds. Try seek for "capacity management". Depending on how large the environment is, but the crucial part is always to talk with the client about what's his expectations are. As consultant it's your mission to fulfill those expectations first.

Level 18

The tool has merit...but is based on historical data.  Granted some extrapolation of data may represent growth and increased load, but there are always other variables that cannot be accounted for especially if you are adding new applications and servers into the mix.

Level 10

Jfrazier I agree.  We can make some very good estimates these days, but without always knowing what is on the development horizon, it can be somewhat of a crap shoot.  The tech field has become more dynamic than ever and with that I don't think we will have it down to a perfect science - at least for a while.

Level 12

Totally agree. You can make an educated guess, but that's it. In my experience, trends have a way of spiking as soon as you call them a trend.

Level 11

Nice post

Level 21

I have worked with tools that do this including VMAN and found the workload simulations to only bet partially useful... when they work.  Unfortunately I have found the data from such tools inaccurate a few too many times to trust them completely.

I personally have found just some basic baseline data such as host or cluster performance data combined with knowledge of the environment and applications to be as useful if not more useful for making predictions about what additional workload the environment can handle.   

Level 18

Agreed...but it does give you a starting point.

Level 10

We've had a lot of rapid growth in a short time with one conversion right after the other as we're buying up smaller businesses so its nearly impossible to know what we're going to need in the future. The method that we use though is we have a 3rd party consulting group that we purchase our VNXS's from and they help us with planning for what we need to buy as part of the service.

Level 18

Ahh...rapid growth further compounds the ability for "forcasting/what-if" tools to predict growth as it is seldom linear.

Level 12

On a slightly related note, I like having the ability to right-size virtualization assets.  In the most extreme case of this, I once had a client whose application group insisted a new application VM required 24 GB of RAM to run properly and they built it as such.  However, looking at the historical data, we saw that at no point had this server ever exceeded 1.5 GB of RAM utilization.  Needless to say, they revisited the requirements for this VM and made the appropriate adjustments.

Level 10

Interesting What-If Scenario and tools to assist with predicting the out come.

Level 15

Interesting subject matter.

About the Author
Virtualization blogger and IT engineer, living at Reunion Island (fr). Trying to help others with their journey to all virtual...