My VMware colleague Tristan Tod (@tristantodd) put together an excellent presentation about VDI and related storage technologies. One of my favorite parts of his presentation deck was the “A Brief History of Desktop Storage Architecture”, wich in my personal opinion is extremely valuable for VDI architects and administrators. I have added some personal comments and links but most of the work was done by Tristan. Well done Tristan!
In the beginning…there were physical desktops and everything was on DASD (Direct access storage device).
Physical PCs 2.0
Filers at team and corporate level started to arrive and user data began to land on shared storage.
Physical PCs and Terminal Services
With terminal services came further distribution of storage for desktops in that enterprise applications began to land on shared storage and even server storage.
Early VDI for the really unfortunate
Some early adopters of VDI were not provided with SAN access (storage administrators worried about VDI) and started hosting desktops on DASD at vSphere host level.
Early VDI for the more fortunate
The more fortunate early adopters were able to host desktops in enterprise SAN/NAS in the same way that many vSphere customer had been hosting server workloads.
VDI with basic tiered storage
Basic tiered storage involved moving seldom used, temporary, and non-essential desktop VM data off of “expensive” shared storage and onto host DASD. Many administrators continue to do this today. Refer to Save [VDI] Storage using VM Swap File Host Placement.
VDI with advanced tiered storage
Advanced tiered-storage arrived in VMware View 4.0 and allowed storage tiering of linked clones to different storage platforms (based on capacity and performance needs). Refer to VMware View 4.5 Linked Cloning Operations Explained (Part 1) and VMware View 4.5 Linked Cloning Operations Explained (Part 2).
VDI with advanced tiered storage and persona management
With more advanced tiered storage + persona management, further separation of storage was possible.
VDI with local SSD or Flash
With the introduction of the stateless reference architectures for VMware View, high performance DASD (SSD or flash) could be used for host linked clones. In this way, shared storage is lightly used for access to user data, persona, and enterprise applications. Refer to Use Flash Drives (SSD) for Linked Clones, not Replicas and Offloading Virtual Desktop IOs with Atlantis ILIO: Deep Dive .
VDI with auto-tier SAN / NAS
Some advanced storage solutions arrived that promised to be a auto-tier solution where all VDI components could land on the same storage platform but be moved to appropriate backend tier based on IO characteristics. Refer to EMC FAST Cache effectiveness with VDI. In this category I would also put a number of different technologies such as the scale-out like Nutanix with Fusion IO cards and Nexenta ZFS with ZIL and L2ARC.
VDI with storage features in Software + advanced physical storage
Finally, advanced automatic tiering in hardware and intelligent storage features in software (such as the View Storage Accelerator) have shown great promise in providing a simplified front-end storage architecture. Refer to Understanding CBRC (Content Based Read Cache).
We have certainly not covered all possible storage technologies for VDI but this list is amusing and demonstrate how reference architectures and deployment models have changes over the last few years. It’s all about cycles and we keep repeating them as the technology gets better and prices come down.
This article was first published by Andre Leibovici (@andreleibovici) at myvirtualcloud.net.