• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

ESXi Memory Capacity Planning Question

R

renixinq

Limp Gawd
Joined
Feb 13, 2003
Messages
439
I wanted to get some additional input on memory over-committment/memory utilization monitoring.

At my work we have a 3 host vSphere 5 environment. Currently, we have about 25 VM's consisting mostly of non-mission critical applications. Each host has 144GB of RAM for a total of 432GB. Our current VM's are allocated to roughly 100GB.

I really pushing for us to virtualize basically everything under the sun for obvious reasons and one of the main items I want to convey is that we have PLENTY of capacity when it comes to memory. To do so I want to make sure I'm using the right memory metrics to convey how much memory we can plan on using.

To get what is actively being used I've been using the usage % on the VM and the active kb on the host and then using that as a comparison to the consumed rate and then comparing that to the total capacity. In general the majority of our VM's are below 15% usage and same when comparing active to consumed at the host level. So at any given moment only about 4% of our memory is actively being used (100/432*.15).

Now that said I don't want to be going crazy with memory overcommittment but I think something around a 1.2:1 raitio would be well within reason. So accounting for a single host failure I'm proposing we should be able to plan on allocating about 345GB of memory for our VMs (288GB * 1.2).

Does that sound reasonable? I want to make sure I'm not talking out my ass when I get called on this.

Thanks everyone
 
Edit: Please disregard this. I don't think it answered the original question to a good enough degree.
 
I'm not sure you can really put a static metric like that to it. It depends too much on the machines you're visualizing and how much they are used. For example... if you end up deploying a bunch of citrix servers and they get saturated during the day, then you can easily fall into the bad side of overcommittment and have terrible performance. Also keep in mind the hypervisor needs RAM, and there is overhead for each VM as well.
 
Maybe I didn't make this clear but that's kind of why I'm asking the question. I know our current load is pretty steady because we have month's worth of data and there is little to no variation. The systems I'm pushing to be virtualized are the mission critical systems and will have "heavier" utilization (exchange, finance/hr system, etc.). I'll be doing analysis on the loads the physical servers are exerting currently and planning accordingly.

I'm concerned about knowing how much room I have on our current environment and making sure I can convince those around me that memory over-commitment is a good thing when planned properly. Which is were the question about reading the metrics come into play. I know there's no hard answer here but I'm wanting to make sure I'm reading/using the proper metrics to analyze our current load.

Thanks
 
What you're looking at is very reasonable. You could have a VMware partner do a Capacity Planner to get better info..but you're on the right track. The Active Memory is the one to use....used is too high. And 1.2:1 over commit is very reasonable.
 
You must log in or register to reply here.
Back
Top