I got 99 problems but IOPS ain''t one - Option 2


	Related Articles Today's Enterprise Application Delivery Framework I got 99 problems but IOPS ain't one Goodbye Citrix XenServer - Hello VMware ESX BYOPC does not really work with Citrix Dazzle or Self Store Plugin yet.. iPad and iPhone plus Citrix AGEE and GSLB equals PAIN


	About Lee James Wynne Application, Desktop, Server and User Virtualisation Architect @ News International, I live in Devon. Addicted to Dairy Milk bars and Walkers salt and vinegar crisps I have been lucky enough to have worked with Citrix technology since the introduction of Multiwin and I started on VMware with version 2 of ESX. Lee Wynne's Specialties: Application Delivery, Desktop Delivery / Virtual Infrastructure, Citrix XenApp, Citrix Access Gateway, Citrix XenDesktop, Citrix XenServer, Citrix Presentation Server, Citrix Provisioning Server, Application Delivery, Application Virtualisation, Application Streaming, Application Packaging VMware ESX / ESXi Server, VMware Virtual Center, VMware Backup Consolidator, Citrix Access Gateway, VMware View, VMware Composer, Appsense.


	Virtualization Podcasts GreenBytes and Desktone: On-Premises, Cloud-Managed Virtual Desktops Podcast with Stephen O’Donnell and Danny Allan - May 13, 2013 Episode 178


	DABCC TV-Videos LogicMonitor: SaaS Performance Monitoring Video Podcast with Steve Francis - Episode 44

Note : Use this as concept design, not say you should but the speed has been amazing.

Further explanation

PVS Servers are doubling up as NFS servers
As per diagram 1, vDisks are stored in PVS Ram after 1st boot
Each VM has a separate cache drive (D:\) for page file + events and vdisk cache file
XenServer has both NFS Fusion IO drives attached as NFS shares
VM Cached drives (D:\) are stored on NFS Fusion IO
VM Cache drives are split evenly across both Fusion IO drives (presuming you have your apps split across xenapp servers for redundancy)

To continue reading this article please view it at source here : http://www.applicationdelivery.co.uk/i-got-99-problems-but-iops-aint-one/

White Paper: AdminStudio Suite DataSheet

Sponsored Link: Download Windows user profile migrator utility from sepago, the inventors of Citrix User Profile Manager.

Article Tags

Lee Wynne: Application / Desktop Delivery + Virtual Infrastructure

Citrix Provisioning Services News, Support, and Training Resources

Citrix Support and Training Resources

Citrix XenApp (Presentation Server) News, Support and Training Resources

Citrix XenDesktop (VDI) News, Support, and Training Resources

Citrix XenServer Virtualization News, Support and Training Resources

Virtualization News Resources

Article Comments

posted by Guest - 03/01/2011

Active HA Failover
Have you validated this configuration is HA Active Failover capable? From looking at the overview, it still looks like you're tying the write cache to a specific PVS system, whether it's set to cache on server disk or cache on local HD, in the end, you still have a single point of failure. Unless there's something in the design I'm missing, if you create a XenServer SR and a local Hypervisor VHD disks presented to the VMs for the write cache, you're still in the end tied to a specific PVS server. For example, a VHD created in SR1 is tied to PVS1. If PVS1 goes down, how does that VM failover to PVS2? If I'm missing something obvious, please let me know. I also replied to your first post with some other options to consider for maximum IOPs without sacrificing PVS HA. Thanks, DY

» reply

posted by Guest - 02/13/2011

Really?
Where are you getting this stuff? Are you simply making it up as you go? Everyone knows that Microsoft's NFS Client is a Joke.

» reply

posted by Guest - 02/14/2011
Re: Really?
doesnt have to be MS NFS, this could also be the iSCSI target, or the Fusion IO could be elsewhere. I think what he is putting forward overall is the cost per IOPS when hosting PVS write cache (in various forms) on traditional spindles. Visit your SAN dude, tell them you need 20,000 - 30,000 IOPS to cover the write cache for xenapp, xendesktop (the latter including reboots at log off on over 500 VM's).

see what the calculation/cost turns out to be for a SAS RAID 1+0 array (and how much disk space is wasted), plus would that be isolated to your service? I doubt it.

scott.
» reply

Submit a Comment

DABCC User:
Guest
Subject:
Message:
Image Code: