Originally posted at vcdx133.com.
This post provides a Tech101 breakdown of the Nutanix Virtual Computing Platform. If you have heard the buzz-words “Nutanix”, “Web-Scale” and “Hyper-Convergence” and want to learn more about it, this post is for you.
The Nutanix Virtual Computing Platform (VCP) is a Hyper-Converged Infrastructure solution. Termed the “iPhone” of infrastructure, what is the big deal?
Short answer: Ease of use, fast deployment, linear scaling of performance with capacity and richness of product features.
Why should you care? If you live in the world of monolithic and vertically scaled storage, you may have experienced the initial, massive cost of purchasing your storage solution with a 5 year ROI. Then comes the honey-moon period where you experience amazing performance as workloads are moved to the new array. Well before the 5 year ROI mark, you will probably experience performance degradation as the capacity of the array is consumed. Which then initiates discussions with your storage vendor about how you need to spend another boatload of money adding additional controllers, SSDs and disks to the array.
Furthermore, think about how long it takes for the storage team, the server team and the network team to agree on a particular design for a project and the time it takes to order, deliver, install, configure and test that solution. Have you ever asked yourself, “There must be an easier way to do this?”
Enter Hyper-Convergence, where you can now buy individual blocks of infrastructure at an entry price and quickly scale-out your solution (performance and capacity) one unit at a time as you grow.
What is it? Nutanix has taken the “Web-Scale” infrastructure model of Google and Facebook and made it available to the Enterprise. You buy a Nutanix Block (rack mounted chassis) that contains Nutanix Nodes (hosts), storage, power and cooling. This creates a Nutanix Cluster (Software-Defined Storage), which are comprised of Controller Virtual Machines (CVM).
How does it work? The Controller Virtual Machines connect directly to the physical SCSI Controllers on each node (Pass-Through mode) and the CVMs are configured to create a “Nutanix Distributed File System (NDFS) Cluster” (storage pool) via the 10GbE network. All of the pooled storage is classified as local or remote from the point of view of each CVM and VM data is protected based upon a Replication Factor (RF) policy of two or three, which replaces the concept of RAID. RF-2 means a local and remote copy is maintained and RF-3, a local and two remote copies of said data.
Where is the magic? Since each node has its own Controller VM, SSDs and HDDs, NDFS can provide linear scaling of capacity and performance for locally stored data. When a CVM goes down, a remote CVM will takeover serving data I/O to the local hypervisor. When the local CVM comes online, it resumes service. This fundamental concept is what allows Nutanix to provide a “rolling upgrade” of the Nutanix Operating System (NOS). What is the caveat? During periods of CVM failure, local workloads that access remotely stored data will experience degraded performance until the local CVM is restored.
The major components of the Nutanix VCP solution are:
It comes in a number of OEM Hardware flavours (Dell and Supermicro) with the following uses: