John Dix (of Network World) recently wrote an article entitled “Contending with the virtualization hangover” – which is a great, short read. In this article he makes the assertion that we’ve all jumped on the virtualization bandwidth and gotten a great buzz, but the challenges that have emerged have shown that we lack a true understanding of this new paradigm.
John – I couldn’t agree more with a few of your challenges:
“stranding a lot of storage as resources are dedicated to virtual machines”
“difficulties getting a real inventory of VMs and managing them through their life cycle”
“makes it hard to back up the resources and accompanying data”
Of course, I’m largely looking at this from a storage angle. It is clear (to me) that virtualization is not only increasing the demands on the traditional storage environment, but that in order to reach that scale with virtualized environments that we all imagine, we’re going to have to find a new storage paradigm.
I’ve written about this before, but let me be clear – there is no reason why storage administrators need to settle for over-provisioning storage, wasting capacity, rebalancing LUNs/volumes, or increasing their management costs in a virtualized environment. Virtualization (which has been around for a lot longer than the leading hypervisors) eliminates the constraints of a physical computing environment. Due to the hypervisor model and the intracies of virtual networks, security, and the fact that administrators are still managing operating systems I would argue that virtualization is not the holy grail. Ideally, I would like one fault-tolerant, highly scalable system (made of commodity parts), which I can continue to add more applications to without increasing my management costs. Microsoft, Citrix, and VMware are all racing to provide something to live up to this ideal.
Isilon is delivering this for storage today – we’ve taken exactly this approach with OneFS. Instead of building a virtualization layer, a namespace aggregator, or other mechanisms which let you party but leave you with a hangover we designed the ultimate system from the start:
A single, highly scalable, high performance storage system (made with commodity parts) that doesn’t require additional management to scale. Start with 10 virtual machines and scale to 18,000. Offer different levels of security and protection on a per-VM basis and do so while continuing to ride the cost/commodity curve and protect your investment.
You can party without a hangover…