At the VMworld I had the opportunity to talk with the Lefthand guys about their products. I hardly see these products in the Netherlands but read some great reviews about their products. One of their products that really caught my attention was and is their Virtual San Appliance or VSA in short. I received a demo copy at VMworld and have been playing with it for a couple of days now. Although one could argue about the performance, the costs is definitely one of the big pros. Especially smaller companies with just 2 ESX hosts could benefit from this. Also for a test environment or demo environment this is a great and cheap solution.
For those who never looked into VSA, what is it exactly?
VSA is a virtual machine which runs on your ESX hosts. The ESX hosts only need local storage, which is currently limited to 2TB. The VSA VM has a small disk at first but you will need to add a large disk, preferably the rest of the storage on your system. This vmdk file will be turned into an iSCSI target within the VSA VM. But not only will it be an iSCSI target, when you setup VSA correctly you can also mirror the disk to the second VSA VM. This means that with only 2 hosts and 4TB of local disks it’s possible to setup a redundant iSCSI “SAN” without the costs of an expensive device. This iSCSI target will be connected to both ESX hosts and formatted with VMFS. This is why you will lose performance:
VMFS -> VMDK -> iSCSI -> VMFS -> VMDK -> Guest VM Filesystem
Do I have any evidence to prove this? Well I did not do any performance tests yet, if I get the chance I will try to run some. But I guess most of you that will be looking into this product wouldn’t mind a performance degradation, considering the costs.
The cool thing about VSA is that you can take advantage of a wide range of enterprise-class features, such as snapshots, thin provisioning, synchronous and asynchronous replication, automatic failover and failback, and management through a simple and easy to use management interface. It’s even possible to use this VSA VM for DR purposes. Creating a DR site without an additional SAN is possible with Lefthand’s products. SAN/iQ stripes your data across Virtual SAN nodes using their patented SAN/iQ Network RAID capability.
Requirements:
- 1GB of RAM reserved
- 1 virtual CPU with 2GHz reserved
- 5GB to 2TB of disk space
- A dedicated Gigabit virtual switch
Side note: Although it’s a really cheap solution it is something that can be rebuild using only opensource tools, DRBD for instance can be used for replicating data between a primary and secondary host. Although DRBD will not initiate a fail over, Heartbeat will. Combine these two with Openfiler or IET and you have your own diy VSA VM. Although this would be the ultimate cheap solution, it will be hard to maintain and or get some sort of support. But for demo purposes this might be a cheap solution.
Mikhail says
May be we may use
“RDM -> iSCSI -> Guest VM Filesystem”, not “VMFS -> VMDK -> iSCSI -> VMFS -> VMDK -> Guest VM Filesystem” ?
Duncan Epping says
Than it would still be:
RDM -> iSCSI -> VMFS -> VMDK -> Guest VMFile System
would save a lil indeed, but would also mean you need two seperate lun’s. one for the first disk of the VSA VM and one for the second.
Andrei says
Does not look like you can add local storage as an RDM to a VM, so it must be VMFS. On the other hand, something very similar to VSA can be created with Open Solaris and HA-ZFS, minus the management tools.