Why Composable Storage is the Natural Datacenter Strategy for Cloud-Scale Applications

 

Yesterday we announced Kaminario K2.N and FlexTM.   For me, this is the most exciting product announcement Kaminario has made in my 9 years with the company.

 

Fully Converged NVMe Backend

In short, K2.N is a software-defined scale-out array architecture running on fully-converged NVMe back end – including NVMe over fabric networking, NVMe JBOF (Just a bunch of Flash), and NVMe NAND flash drives.  K2.N runs Kaminario VisionOS – the same operating system powering our flagship all-flash array, the Kaminario K2.  Besides NVMe performance, the K2.N takes our concept of scale-up-and-out one step further.  Customers can implement any number of controllers (we call them c.nodes) and any number of SSD shelves (we call those m.nodes) to assemble an array optimized for any level of performance and capacity efficiency desired.

 

K2.N is a radical departure from the traditional storage array concept.  But when you layer Kaminario Flex on K2.N, it is not even fair to call it a storage array.  Flex is a management and orchestration platform that allows storage managers to dynamically compose “virtual arrays” form K2.N resources.  Since K2.N m.nodes and c.nodes are all connected via a low-latency converged NVMeF mesh, the entire resource pool is logically proximate no matter the physical layout.   This simple fact allows Flex to define virtual arrays with no physical reconfigurations.  Virtual private arrays retain all the characteristics of traditional enterprise-class AFAs – High availability, RAID efficiency, performance, etc…   This is composable storage.

 

This is the start of a new era for the storage industry.

 

The VisionOS software architecture that Kaminario K2 is based on has always supported true disaggregation of compute and capacity – this is what enables our highly flexible active-active, scale up and out approach.  The NVMe protocol sets our architecture free.

 

Kaminario K2.N with NVMe-oF and node-level granularity

 

Before I go on, I want to emphasize that our existing K2 platform (that uses SATA drives, SAS interconnects between controller and drive shelves and Infiniband networking between clustered controllers) is the most capable all flash array in the industry (Check out the recent Gartner Critical Capabilities Report – we’re rated highest for the most important use cases).  We see the K2 platform as staying the price-performance leader in the all-flash industry over the next 18 months or so as the premium on NVMe drives erodes.  Since K2 and K2.N run on the same VisionOS software, there will be a very smooth transition to NVMe on a timeframe that makes the most sense for our customers’ business.

 

Why Composable Storage Infrastructure will Dominate in the World of Cloud-Scale Applications.

There are 4 important forces that are reshaping the storage market that position composable storage as the natural successor to classic storage arrays.

 

  1. The IT world is steadily moving toward as-a-service infrastructures. Josh Epstein has written extensively on the SaaSification of the enterprise and the rapid displacement of traditional on premises hosting of enterprise applications.  Beyond SaaS, we see the world of on-demand applications growing faster than traditional enterprise IT.  From Fintech to Healthech to eCommerce to IoT, there is an enormous growth in as-a-service infrastructure.  IT organizations supporting these environments need to think differently about how they procure, manage, optimize, and orchestrate storage resources.  Composable storage addresses this need.

 

  1. There is an unmet need for storage solutions that deliver the operational simplicity of public IaaS AND the performance and efficiency of shared storage arrays. Composable storage allows IT organizations to get the performance and efficiency at scale, while keeping management of storage resources simple.   By integrating composable storage platforms with higher level data orchestration platforms (like Chef and Puppet), storage operations can get extremely efficient.   (Kaminario Flex will include an open set of APIs for integrating to other datacenter software management).

 

  1. Modern applications need to be fast. And they incorporate sophisticated real-time analytics.  Kaminario has always been focused on the ability to deliver cost-effective performance in analytics intensive environments.   N takes this capability further by bringing increases in raw performance and performance density AND the ability to layer on as much controller compute as needed to reach a desired throughput.    It might be crazy to build an array with 10 controller nodes and 1 JBOF.  But it is simple to temporarily allocate 10 controllers from underutilized arrays to augment the surge processing needs required to respond to a market event.

 

  1. The distance between the application layer and the storage layer is compressing. Development and Devops teams want to take advantage of storage capabilities to optimize the performance, capabilities and user experience of their applications. The concept of a physical array does not align to the way modern application are being designed. New applications have to “expect the unexpected” and adapt to rapid changes in their environment. The infrastructure has to align.

 

This is the start of a new era for the storage industry.  I suspect we will hear from other established players and new startups that push the idea.  For Kaminario, this is the natural evolution of the storage architecture we began building 9 years ago.   We see enormous opportunity to create value for our customers and our partners.

 

New Call-to-action