Welcome: To stay updated with all my Blog posts follow me on Twitter @arunpande !
In this blog post I will discuss some basic sizing guidelines which you should consider when sizing NetApp storage for VMware View environment. The success of the VDI infrastructure strongly depends on the how the compute & storage is sized & designed to meet the user requirements.
As a Virtualization Consultant in my previous role I have experienced that in one of the projects the Storage was sized only by considering the following factors:
Number of IOPS per user
Number of users
Total IOPS = Number of IOPS per user * Number of users
Really?? I was surprised with this approach but couldn't comment since I was working as a Server Virtualization consultant and the Storage guys felt that their approach is correct. I would like to share the additional points that you should consider when designing Storage for View. You can also use these points as questions during the interviews with the stakeholders.
User Type: The user requirements would vary in a VDI environment. There may be few power users with very high IOPS requirement whereas there may also be users who need very low IOPS. Also consider how many concurrent users would be logged in and the different shifts they would work. This will help you size storage more efficiently.
Desktops Type: This is a very important point which is left out sometimes in the initial discussions. The Storage requirements for Persistent & Non-Persistent desktops are different. For e.g. the non-persistent desktops have to be deleted and re-created at regular intervals. When you have existing groups of users already logged in the deletion & creation of desktops would require additional IO, CPU & Disk usage.
Protocol Type: The protocol choice is not only restricted to FC, iSCSI & NFS but you should also consider CIFS or SMB for user profiles.
OS disk size in GB: Always size the C as per the OS requirement. This will ensure that you get maximum storage space savings through dedupe and reduced IOPS using other caching technologies. For user data, persistent or non-persistent use a data disk. In general disposable disks are created on low cost SATA disks whereas persistent disks are created on SAS disks which require comparatively more IOPS.
Unique Data per VM: This parameter will affect factors like dedupe and in turn make considerable changes in sizing the storage.
Memory size in GB: It’s important that you size the VM memory appropriately because the swap disk created on the datastore for each VM would be of the same size. Size the VM memory as per the OS & application requirements. Also consider memory savings using VMware TPS and size accordingly.
Read/Write Ratio: Here one has to understand the ratio of Read vs Write IOPS of the total IOPS per user. For example, certain users may require 50/50 however other users may require 20/80. This will help you in sizing the Flash Cache if the workload is read intensive or SAS/FC/SSD disks if the workload is write intensive.
Read/Write I/O size: This may vary depending on the requirements from 4KB to 16KB. This will determine if there would be large number of small read/write or small number of large read/writes.
NetApp Storage: In addition to sizing storage also consider some headroom for aggregate/volume snapshots.
Cloning Method: You may also consider using NetApp Rapid clones when creating full desktops on NFS datastores. This will result in faster provisioning time and reduced Storage capacity.
Guidelines for Virtual Desktop Storage Profiling and Sizing
Storage Considerations for VMware Horizon View 5.2