VMware has announced that they have entered in a definitive agreement to acquire Virsto, a vendor who offers a “storage hypervisor” for virtualized environments. This is likely to factor significantly into VMware’s SDDC strategies.
Virsto and the VMware SDDC
We all know that VMware has successfully virtualized the CPU and memory resources in servers, abstracting the management of these resources away from the hardware and putting them into a layer of software (the vSphere hypervisor, ESXi). We know that VMware delivered VXLAN in vSphere 5.1 as the first part of its network virtualization strategy. Acquiring Nicira was certainly part of the further implementation of that strategy.
But when it comes to storage, VMware has had relatively little to say until now. Clearly if you are going to have a Software Defined Data Center, in which all of the resources in the data center are virtualized, that has to include storage. To get an idea of what might be possible, read the I Have a Dream blog post by Mark Davis the CEO of Virsto.
To really understand what might be possible with the integration of Virsto’s technology into vSphere, have a look at the blog post by Comac Hogan that goes through how the technology works and its benefits in detail. The short answer is that you save on storage utilization (a lot), you improve read latency by 30/40% and you can improve write latency over a SAN by a factor of up to 10X. This is possible because:
- Virsto consists of two pieces, a service that runs in the vSphere host, and an appliance which sits in-line in between the hosts and the storage arrays.
- Virsto therefore has complete control of the protocol from its appliance to the disks, how data is written to those disks, how things are cached, etc.
- This allows, for example, in the case of writes for Virsto to log writes to a fast storage device (like SSD) and then change the order in which those writes are actually written to the disk. So if a database is doing sequential writes, but the I/O blender in the hypervisor makes these writes into random I/O, Virsto can reorder the writes to make them sequential again.
- Virsto does all of this management of the back end storage in a manner seamless to the VMware adminstrator. To the admin and to the vSphere hosts the entire pool storage managed by Virsto just looks like a NFS datastore.
In summary, with Virsto, VMware can now completely abstract the physical storage from how it is presented to the servers and the applications on those servers (the storage is virtualized), and have fine grained control over how data is read and written from the datastores – resulting in dramatic improvements in utilization and performance.
Share this Article:
Latest posts by Bernd Harzog (see all)
- VMware vSphere 6 Attacks Amazon with “One Cloud, Any Application” - February 9, 2015
- VMware vSphere 6 Attacks Red Hat: VMware Integrated OpenStack - February 3, 2015
- Will the Public Cloud Be the Next Legacy Platform? - January 20, 2015