VMware Cloud Community
jhable
Contributor
Contributor

Low energy diskless server set up with iSCSI

Hi, I currently have a highspec supermicro with 2 x QUAD core processor, 8gb memory and areca 1220 raid controller and esx3.5i. It is very loud and expensive to run plus I get the odd purple error screen which I need to investigate but think it is disk/controller related.

I want to change my setup to use 1 or more low energy diskless servers running esx4i and have all the vm's and data running on an iSCSI device such as ReadyNAS 2100. I would like to detach the disks/controller from the esx4i setup so that I can load esx4i on a USB stick and just plug it into another server if one should fail for instance or I need to create more vm's and the low cost server is struggling.

I am looking for suggestions for the low energy diskless server setup and currently use the following vm's. windows 2003 server as PDC, windows 2003 server with exchange, windows 2003 server with SQL, windows XP with VS2003 as development, windows 7 with VS2008 and SQLExpress 2008 as development. I am slowly phasing out the xp and sql vm's but need the facility to create test/development environments as and when.

The ups and server I currently have are rackmounted and ideally this is what I would like but I want to reduce noise and power consumption. Any suggestions would be appreciated. Although it is a corporate type setup, it is more for support/testing as there are only a couple of users.

Thanks in advance for any advice.

Reply
0 Kudos
7 Replies
J1mbo
Virtuoso
Virtuoso

That list of VMs sounds like a lot for an 8GB machine. What power budget do you have overall?

Please award points to any useful answer.

Reply
0 Kudos
jhable
Contributor
Contributor

I don't have a specific budget. The vm's run quite easily on the supermicro at the moment but bear in mind this is more a development/test environment. There is mainly one user on the system so although it is networked there isn't a great deal of load on it.

If I look at the stats on vmware the processors are hardly being touched and the memory is just about okay although any more vm's and it will be pushing it a bit.

What I really want is to cut down the noise, have more resilience and cut the energy costs if possible. I don't believe I need the 2 x QUAD core xeon's and I would like to seperate the disk access from the vm's so that if for instance I have a setup that can't cope with the number of vm's I can just buy a low cost diskless server and run extra vm's. If I don't need some vm's for a while then I can shut that server down for a bit or if a server fails I can just fire another one up with either embedded esx4i or esx4i on USB and continue to operate.

Reply
0 Kudos
J1mbo
Virtuoso
Virtuoso

Well it depends on your capital budget. Most likely you could save quite a bit of power consumption by removing one of the CPUs that isn't required and replacing the PSU for an 80+ model, if it isn't that already. The stats don't surprise me, upping the RAM to 16GB if possible would be a good start.

I can't commend on the NAS device as I haven't used it, sorry. You might also consider StarWind or Drobo - but whatever, it seems to me that you'd just be moving your single point of failure to the storage device.

Please award points to any useful answer.

Reply
0 Kudos
jhable
Contributor
Contributor

Thanks for the reply much appreciated but I feel I am not getting my point accross. I am aware I could remove a processor and did consider it but that would only be a part solution. I have got redundant power supplys as well in the supermicro but that wouldn't be important if I effectively had redundant servers. I realise I am moving the point of failure to the iSCSI but I also have another NAS which could take over if neccesary.

When I initially installed esxi it was a bit of a pain because the Areca controller was not natively supported and I had to download the driver from Areca and then mess around a bit to get it installed etc. The idea is that I can ignore any disk controller device drivers and have a bootable USB stick that I can boot up a 64bit server with and then connect to the vm's and away I go.

Reply
0 Kudos
J1mbo
Virtuoso
Virtuoso

It seems to me the requirements are somewhat in conflict - adding external storage is just bound to increase the overall power consumption.

I would personally suggest using only hardware on the HCL, which probably explains the issues with the controller. ESXi is supposed to run "as is", with nothing added or changed.

Hopefully someone else can offer some input on the NAS etc.

Cheers

Reply
0 Kudos
jhable
Contributor
Contributor

Thanks. Just for info i think the NAS is only about 80 watts and the

server was in use before I decided to use Esxi and I did not want to

spend more money on another controller at the time.

Reply
0 Kudos
Rumple
Virtuoso
Virtuoso

You won't be able to use mainly diskless servers unless you have ISCSI HBA's and can present the LUN's correctly. The cost of the ISCSI HBA's will be more then just runnng a full blown server. Hell I have an ibm x3650m2 with 2x300gb drives, 48gb of ram and 2xquad core processors that was under 12k CDN...Scale that down and put in smaller drives and you can do it for 5-6k.

The problem with the ISCSI NAS is that you will be sorely disappointed if you go not get a NAS that has a BBWC on the RAID controller. Most, of not all the low end ISCSI NAS devices (all the ones I found to date below $2500usd) are using an Intel controller that is hardware RAID without the BBWC capabilities.

Trying to run multiple servers off non cached storage is going to not run well and not make you happen when you realize the money could have been better spend on a single server with local storage...especialyl for a smaller environment.

On a side note...the 3650m2 was amaxingly quiet...I run it currently in low power mode and I honestly could not believe how quiet it was when I was setting it up in my office before going to the datacenter with it. My cisco 3750 swithc was louder...

Reply
0 Kudos