Q&A with Tiffany To of Nutanix
VSM: Nutanix has been messaging virtualization without a SAN. Why do you believe this is a better approach?
TT: No one disputes that the reach of virtualization is greater than ever before, but in terms of penetration in a single organization, many non-Fortune 500 companies are not yet moving beyond limited server consolidation. Feedback from VMware customers points to the fact that many customers feel they aren’t getting the ROI they expect on many virtualization projects due to the cost, complexity and performance management of their SAN/NAS. Compute power has scaled and now SSDs enable massive IOPS, but this performance is throttled by the network, creating more expensive and complex configurations over time. At Nutanix, we feel it’s time to completely rethink the infrastructure and merge enterprise storage with x86 servers into a single tier for a converged, scale-out virtualization appliance.
VSM: What exactly is a converged virtualization appliance? What does the Nutanix solution look like?
TT: We believe that customers who want a simpler way to virtualize are going to challenge the current server and SAN/NAS approaches. The big vendors recognize this and have released prepackaged solutions that simply repurpose existing products without tackling the bigger architectural challenge. If you stop to consider the excess CPU power on existing storage solutions, it’s not much of a stretch to think about the benefit of simply bringing the virtualization host and storage together for better performance. The big challenge, though, is how this type of architecture can scale out. Building a distributed file system that can manage this type of appliance architecture as you grow from a handful of nodes to hundreds or thousands of nodes requires true innovation.
The Nutanix Complete Cluster delivers a complete virtualized datacenter in a compact 2U building block that houses four servers which each act as a virtualization host, preloaded with vSphere 5, along with storage for those virtual machines in the form of 20TB of SATA drives and four Fusion-io ioDrives. The storage across each node in a Nutanix Cluster is logically aggregated so it can be sliced and diced as needed for your virtualization workloads, and a software-based controller VM ensures that performance scales linearly as more nodes are added, unlike hardware controllers in traditional storage arrays.
VSM: Is this like a Virtual Storage Appliance (VSA), e.g. Lefthand or VMware’s recent vSphere Storage Appliance? How is Nutanix different?
TT: VSAs weren’t designed for enterprise virtualization, and can’t deliver high performance or scale out to more than a handful of nodes. These products work for certain use cases in branch offices, for example, but aren’t really a fit for enterprise workloads that demand features like high availability. The Nutanix Complete Cluster is designed for high-performance, enterprise virtualization with features like high availability, snapshots, disaster recovery, etc. and scales-out from a few nodes to a few hundred nodes. We recently validated a 52 node configuration named Project Colossus that demonstrated 375,000 random write IOPS and 224 Gbps of sequential read bandwidth.