Hi guys,
I am planning on building a Fileserver on my ESXi 5.1 and just want to make sure I am going the right route. Hope you can give me an advise here.
ESXi:
5.1 (Build 799733)
2x Intel Xeon E5606
Supermicro X8DT3 Board
48 GB RAM
QLogic QLE4062C
Virtual Machine:
Windows Server 2008 R2 SP1
8 vCPU
4096 MB RAM
stored at DAS RAID6
The data (Office files, Adobe Photoshop, InDesign,etc) is located in an iSCSI SAN. There is one target connected to the ESXi host with 12 LUN (smallest 50 GB, biggest 650 GB). The volumes are attached to the VM as pRDM.
Is this a good way to do this? Or would you recommend me a different way?
If you need more information, please let me know.
Thanks in advance!
Welcome to the Community - I do not think you will need 8 vCPUs ofr your VM - I would start with 1 or 2 at the most - everything else looks good -
Welcome to the Community - I do not think you will need 8 vCPUs ofr your VM - I would start with 1 or 2 at the most - everything else looks good -
It feels kinda strange to have 12 pRDM disks attached to one server. I am a bit afraid about having a bottleneck here. The ESXi hosts is connected trough 1 Gbit with the iSCSI storage.
Does is make a difference how many disks you have connected to your VM performance wise?
Hey,
I'm sure at somepoint there will be some limitation factors, but with the specs you have outlined I do not believe you will have an issue. We have a similiar setup,but for a server running Backup Exec 2010 which has 10 seperate disk connected to it with (1) Vnic at 10gb. I only have 8gb of ram and 2 VcpUs. Everything works as expected running 8 and sometimes 10 backup jobs simultaneously. Are all of your iscsi luns in the same Vlan or multiple Vlans and are they coming back to one controller or two?
Regards
All iSCSI LUN are in the same VLAN. And there are coming back to one controller, if I understand you correctly. No MPIO or similiar.
Are Iscsi luns 1gb or 10gb currently, and would you happen to know the amount of cache on your controller? Also check if there are any input/outdrops on the Core switch the current IScsi Vlans are currently coming back to? If its a Cisco running IOS device you can issue a
>show interface gigx/x
or
>show interface tengigx/x
Regards
The LUN are connected via 1 Gbit. The storage controller is an ARC-1680 with 512 MB cache. I've checked the switches and there are no errors at all.
Start with half the luns, test, check the results and add the remaining if no errors occur.
Why should I reduce the disks? How do you test (performance)?
Reducing the number of disk would give you some sort of baseline. Originally we only had 4RDM and then we noticed if you run 'X' backup jobs all at high priority for key services such as SQL,Oracle,and Exchange,etc coming back to the same 1 RDM we had peformance issue. At the time the other 3 RDM were backing up additional services GRT on AD, file servers,financial servers,web servers,etc. Afterwards I increased the RDM from 4 to up to the number we have now and we have not had any performance issues. Additionally I have rezoned some of the paths because we use FCOE to only go to certain controllers and certain vsans. In your case or any for that matter, its always a good idea to create some sort of baseline and increase as required. The same goes for the server built for this project. You never know, you may require those 8 proc and 48 gb of ram elsewhere in the infastructure.
Is it the same performance wise, if you have 1 target with 10 LUN compared to 10 targets with 1 LUN each?
I'm not sure what you mean when you state 10 targets. Can you clarify?
A good way to consider luns are like highways. The more paths "highways" to disk the better your overall performance will be within those luns. If you are planning to have multiple 10 (I/O) transaction occuring at the same time, I would recommend going with 10 luns, but keep in mind there is a max limit on ESX/ESXi.Hope this answers your inquiry
Regards
Well, you could create a target for each LUN. Or you could create one target and add all LUN to it.