2

We have (gradually) built up a simple network of servers which I intend to use as a virtualisation stack(?). To illustrate the hardware that we have at our disposal please see the following image:

Network Layout As you can see we have three rack servers that I intend to use as ESX hypervisor servers (two out of three are already installed and running). Each of these have two 73GB hard drives installed.

We have two DL380 G3's which currently have 4 73GB hard drives between them. I am more than happy spending money increasing the number/size of the hard drives in the two DL380's.

We also have an ML110 G6 which is currently serving files to users on the network and also serves as the vCenter appliance.

Ideally I would like to use the keep the current configuration but move the files to a virtual machine or somehow serve them from one of the SAN's.

My question(s)

  1. If you had the same hardware as we do, would you set them up any differently?
  2. What RAID setup would you use for the DL360's running ESX?
  3. What HDD and RAID setup would you use for the DL380's?
  4. How would you make the DL380's a SAN? What OS would you recommend?

For additional information the servers have the following specs:

DL360's: 2x 3.0GHz Xeon Processors with 4GB RAM. 2 x 73GB HDD's

DL380's: 2x 2.3GHz Processors with 3GB RAM. 4 x 73GB HDD's between them

ML110:

2x 2.8GHz Processors with 8GB RAM. 1 x 250GB HDD

Improve this question
8
  • Please specify the generations of your Proliants. Based on the CPU specs, I strongly suspect they're too old to run ESXi properly. And based on your other comment I'm pretty certain you'll need at least two new switches. And you should also specify what NICs you have in these servers. Don't expect to get away with just the built-in pair of ports.
    – Max Alginin
    Jul 9, 2011 at 20:56
  • The DL360's are G4's, the DL380's are G3's and the ML110 is a G6. They are all using the built in NIC's. The DL380's have two fibre ports each.
    – dannymcc
    Jul 9, 2011 at 21:11
  • Yup. You're out of luck. You won't be able to run 64-bit VMs, they're not on ESX(i) 4 HCL, and in general those CPUs are just not good enough. This doesn't mean you can't make things work with that hardware, it's just seems somewhat pointless: you'll sink a lot of effort into something that will never give you good results and will never be supported by VMware.
    – Max Alginin
    Jul 10, 2011 at 3:41
  • If they're not on the HCL does that mean I can't run two Windows Server 2008 vm's alongside maybe four other Linux VM's?
    – dannymcc
    Jul 10, 2011 at 8:37
  • That means you'll never get any support from VMware. Every time your ESXi host PSODs, you won't be able to do anything other than reboot and hope it doesn't happen often. And think again about the no-64-bit-guest issue. Can you guarantee all your current and future guests will be 32-bit? I strongly believe that at this point your best move would be asking another question on Serverfault, something like "what's the cheapest reliable hardware platform for ESXi 4.1?" I wish I could help here, but we're a single vendor (Dell) shop, and I don't know much about white box alternatives.
    – Max Alginin
    Jul 11, 2011 at 15:51

1 Answer 1

Reset to default
3
  1. You seem to have the general idea down. However, how are you handleing the switching infrastructure for storage? How are you going to handle data networking for everything else? I.e. VLANS, link aggregation ports?
  2. If you have no vmdks sitting on the hw, raid 1 is sufficient, as the only thing you are worried about is losing the o.s.
  3. It depends on the o.s. you choose. If you use solaris/nexenta as I state in the next answer, you may look at a raidz2 or mirrored. With the amount of spindles you have, you are not going to get that great of performance.
  4. I personally would look at either solaris 11 or nexenta as the software to run on the DL380s. This will allow you to build a decent iSCSI SAN. Look into the limitations/costs of both. This is only if you feel confortable with *nix based systems, as one wrong move can cause serious downtime.

Issues to look into:

  • Take a look at xen cloud platform, just so that you are making an informed decision. (If you already have esxi, then skip this.)
  • Do you have enough memory for what you are trying to serve?
  • Look into the number of spindles on your SAN
  • Look into your switching/networking infrastructure.
  • On the storage side, only use gigabit switches and make sure they support jumbo frames.

If you have questions let me know.

Improve this answer
1
  • Thanks! To be honest VLANS and link aggregation completely baffle me, but I guess that means that this is a great learning project. The servers all connect via a switch that only has two gigabit ports. If it makes a difference the two DL380's seem to have fibre ports in the back.
    – dannymcc
    Jul 9, 2011 at 17:10

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .