Citrix MCS an overhead on IOPS ?

During some testing of some flash based storage i noticed something strange in my test results. 

Machines created with MCS (Machine Creation Services) seemed to not be able to perform as well as machines cloned full fat by the hypervisor. 

Conducting testing with IOMeter setup with a VDI type workload as detailed in another blog article we saw the IOPS saw the following results.

MCS Cloned Machine Peak IOPS 3500

VMware Cloned Machine Peak IOPS 40,000

This was using the same master image, on the same VLAN, on the same Hypervisor host. 

The MCS created machine was a dedicated VM type. 

The VM was assigned 3GB of RAM , 2vCPU and the disk size was 80GB

The storage was isolated for the tests. 

The test was repeated over a number of times to validate the result.

I understood that using linked cloning would have a slight overhead over a full fat VMDK file but did not realise the overhead could be as high as 10X

It should be noted that 3500 IOPS per desktop is still good enough to deliver a great experience to users, and the feature benefit of pooled.

IOMeter IOPS Benchmark Testing

The freeware tool IOMeter is very handy for testing the kind of IOPS that a given bit of storage can deliver.

The following are some recommended settings for using IOMETER to test IOPS for a VDI solution.


Atlantis computing have a newer article and set of setting detailed here.
Atlantis Advised Settings

My testing has shown best results with 6 workers , disk size of 2071952 and 16 threads per worker.

A full guide to IOMeter can be found here

http://www.randylee.com/hp/iometer_images/Iometer.pdf

A good guide to IOPS testing has been written by Atlantis Computing
Windows_7_IOPS_for_VDI_a_Deep_Dive_1_0

Citrix Storage Basic information

The following is some summary information for the citrix storage guide.
CTX118397


Hard Disks SATA v SAS

Solid States Disks (SSD)

Bus Technology Speeds

Citrix recommendations

Storage architectures

DAS - SATA/SAS/Fibre channel

SAN - block level -  NFS/CIFS/SIFS

NAS - file level -  HBA Card connections

NFS

Fibre Channel

iScsi

Citrix PVS recommendations

Citrix Performance recommendations

Disk Raid Levels

Multipath

Multipath I/O is a fault-tolerance and performance enhancement technique. ―Multipathing
solutions use redundant physical path components — adapters, cables, and switches — to create
logical paths between the server and the storage device. In the event that one or more of these
components fails, causing the path to fail, multipathing logic uses an alternate path for I/O so
that applications can still access their data. Each network interface card (in the iSCSI case) or
HBA should be connected by using redundant switch infrastructures to provide continued access
to storage in the event of a failure in a storage fabric component