All Nutanix products were designed from the ground up with performance and scale as the two driving principles. Data typically passes through the local OpLog with a copy sent over the network to another node in the cluster for redundancy. Sequential writes skip the OpLog and go directly to disk, and they may optionally skip the SSD tier entirely for specific use cases. This reduces the amount of storage needed on the SSD tier while taking advantage of the suitability for sequential writes of HDDs.
Measuring performance on a Nutanix box is something Nutanix does as a matter of course. Since the movement of data to and from the underlying storage is completely controlled by NDFS, it's also possible to monitor and track the moving parts in order to identify any bottlenecks. This snapshot of the vDisk status page shows the different types of performance measurements available.
Nutanix provides a diagnostics tool that will provision a VM per node with six virtual disks attached. Once the VM has been fully provisioned, the tool launches diskperf and fiotool to measure various performance parameters under known loads. Once complete, the results are aggregated to determine the overall cluster performance. A typical user won't run these tools, but they are available for the Nutanix System Engineers to use as a part of the post-installation process.
The Nutanix NX-3000 series of products provide a unique solution for virtualization deployments. Measuring this product against any other competition would be problematic as there really isn't anything like it. It's even tougher if you evaluate strictly on cost, which at a base price of $144,000 per appliance is significant. But the benefits — high availability, high performance, all the advantages of centralized storage without the overhead — are compelling. At the end of the day it's an ideal solution for high-end data centers looking to pack as much virtualization capacity into the least amount of space possible.
Sign up for MIS Asia eNewsletters.