VMware Cloud Community
paulnc2008
Contributor
Contributor
Jump to solution

Problem configuring an HP Proliant.... How much memory is enough?

Thank you for reading;

I have used VirtualBox, QEMU, Virtual PC, and others for some time to virtualize instances of various Linux and Windows Guest OSs.  In the company I for which currently work, there are multiple Proliant DL380s and the vmware used on them (though dated) is very functional.  I am having problems comnvincing some internal people that we just do not have enough physical memory on the systems to  effectively grown avirtual network to satisfy our testing needs.  For example, the construction/implementation of 15 Linux systems on one server took almost a week.  I have tried several ways to speed things up (the creation of the appliances) but to little avail. (i.e. I have exported an appliance and then gone through the lesser pain of importing that appliance several times, done vanilla installs simultaneously, etc...)

The proliant in question has 12GB of memory, 2X Intel E5504 processors, 2 500GB volumes, etc.

SO, the basic questyion is:

How do I speed the creation and installation of VMs?  What Guidelines should I use to upgrade the Proliant's memory?  WHat is the overhead of vmware itself on the proliant?  I assume with that many VMs running, there is going to be a LOT of swapping, just due to the lack of physical memory.  ANY leads will help.

Thanks All

0 Kudos
1 Solution

Accepted Solutions
AndreTheGiant
Immortal
Immortal
Jump to solution

Welcome to the community.

I suggest to read some performance graphs (mem, cpu, disks, ...) to see there could be the bottleneck.

Usually, with local storage, the problem is on disks parts (be sure to have a latency less than 20ms).

Andre

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro

View solution in original post

0 Kudos
6 Replies
logiboy123
Expert
Expert
Jump to solution

Use templates to speed up creation and deployment of VM's or simply clone the VM's that you have. Guest customization is possible on all Windows and most corporate versions of Linux.

How do you know that you need more memory? If all your ESXi hosts are running at maxed capacity on memory then find out how many memory slots are available on your motherboard and add some sticks. I would stay with using the same memory module through a single server. So it may be worth moving some of the RAM in one host to another and then fully populating that host with all new RAM.

There is about 1 GB of memory used per ESXi host. This is a rough guide though, it depends on your environment.

Generally speaking systems page when they run out of memory. Design against this and you don't have to worry about it. But having fast storage is always helpful.

Cheers,

Paul

0 Kudos
idle-jam
Immortal
Immortal
Jump to solution

i also depends how much virtual machine with assigned virtual memory. although there are memory optimization technique like transparent page sharing, balllooning and etc, i will just play safe and use the sum of all virtual memory as a guideline.

SyApps
Contributor
Contributor
Jump to solution

I'm not really convinced that memory is the reason for your slowness when building out new VMs. How many VMs are up and running and how much of the 12GB are you utilizing when you try to create another VM? How are your disks configured? RAID? As it was previously suggested, you might have better luck if you built out a template and deployed new VMs from there. To be honest, you should really have less then 6 VMs running on a box with only 12GB of RAM. This is relative and depends on how much RAM the VMs use.

Always a big thanks to the community in advance! Dan Lee
AndreTheGiant
Immortal
Immortal
Jump to solution

Welcome to the community.

I suggest to read some performance graphs (mem, cpu, disks, ...) to see there could be the bottleneck.

Usually, with local storage, the problem is on disks parts (be sure to have a latency less than 20ms).

Andre

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro
0 Kudos
paulnc2008
Contributor
Contributor
Jump to solution

To elaborate;

The suggestions are fabulous but I wish to make a case for more memory, since we will be running 15-20 VMs on the server for stress testing and simulation.  We typically assign a maximum of 1GB of memory per VM and @20GB/disk.  our installs are usually linux (Ubuntu, SUSE, RHat/Fedora,etc) with a mix of Windows XP and a couple of Win7.   The Proliant server config is:

S/W

VMware ESX Server 3i v3.5.0 accessed via VMware Infrastructure client v2.5.0

Hardware:

2x500GB volumes (mirrored, Raid0)

dual intel Xeon E5504 (4-corex2)

12GB memory

Symptoms observed:

Building a VM utilizing an .iso is SLOW to be sure.  even with no active VMs it takes 2-10 times longer to build an Ubuntu (for example) than to build the same system (exactly) utilizing VirtualBox or even the archaic VirtualPC. (either x386 or x64)

Based on these specifications, how could a case be made for more memory given what the system will be used for?  I know that active systems will make the specifications/analysis tools look like things are under-utilized but with page faulting, etc, things running at a crawl at best. 

I am a n00b to VMware and the proliants.  Help please in making a case for upgrading the memory (or even disk).  also, what memory would be best for what we have?

In advance, thank you for your help, I will keep all informed

0 Kudos
SyApps
Contributor
Contributor
Jump to solution

Well, the good news is that you're answering your own question in a certain sense. If you're hoping to get 15-20 VMs on that server then you're going to need more RAM then 12GB, guaranteed. I stated before that RAM allocation is relative to how much each VM needs, but there's no way your VMs are going to run well on less than a gig. If you take away the hypervisor and just install Windows 7 on metal, can you run efficiently with less than a gig of RAM? Of course the min requirements say it will install, but install and run well are two different things. I wouldn't touch a Win7 machine with anything less than 2GB. I just don't have the patience...

Regarding your situation, it may be better if I gave you an idea of what we run. My oldest environment runs HP, so I'll use that as an example. I'm running a 3 server ESX 3.5 cluster, DL365 G1's with 16GB of RAM and 4 CPU x 2.8 GHz. 16GB isn't a lot, especially when I have 8 servers running WinServ 2003 Standard on each of them, but their usage is very low. They work their hardest when they are running backups at night and their and right now they are using 8.18GB of their 16GB, which is perfect as they don't all need a few more gigs of RAM at times. They spike differently. It's just the nature of the environment.

I also support a different environment that would fail miserably if I had allocated anything less than 6GB per server.

I guess my point is, check your monitors and history and see what you are using. I'd upgrade for sure if your environment was under so much stress that it maxed its RAM usage even once.

Always a big thanks to the community in advance! Dan Lee
0 Kudos