In 2013, the infrastructure industry witnessed a great shift in the way architectures were constructed and deployed. Lead by Nutanix, Simplivity, and VMware, the world of Hyperconverged Infrastructure (HCI) was born. The goal of HCI was to eliminate the complexities and bottlenecks associated with traditional SAN technologies. By converging storage and compute in to a “node” and leveraging state-of-the-art software to join the nodes resources, these pioneers built a new architecture that achieved the “no-SAN” goal.
Fast forward to 2017 and HCI solutions are now mainstream, steadily capturing market share from the larger storage vendors – EMC, NetApp, Hitachi, etc. No longer a niche solution, HCI is now in the process of expanding and growing in accounts and capturing workloads. As the environments have grown, an interesting side effect of the “node” has become apparent – this side effect is what we call “resource drift”.
Resource drift is directly associated with the most precious resource problem. In a node where memory, CPU, and storage are joined together, there is always going to be one resource that is going to be the most precious resource based on the governing use case. As the use case scales, the constraining resource will force the purchase of additional nodes, but because all the resources are locked together in the node, the less constrained resources come along for the ride becoming stranded. As the environment grows, the stranded resource pool grows larger and larger.
For non-technical people, like my spouse, this is a tough topic to conceptualize, but I came up with a great analogy. Let’s tie complex use case like VDI to Neapolitan ice cream. In the case of VDI, we will make memory strawberry ice cream and make CPU and storage vanilla and chocolate respectively. Now VDI is very strawberry heavy. As I add more users, I have to add more strawberry, but because I am using a Neapolitan “node”, I am forced to purchase more vanilla and chocolate, even though I really don’t need it. As the need for strawberry goes up, my freezer continues to have vanilla and chocolate build up. Because I have no real way to consume the other flavors, they just build up causing a massive resource drift over time.
In the real VDI use case, we know of a real-world Service Provider that leverages node-based Hyperconverged technology. His users use a lot of memory and each time he adds users, he also needs to add more memory. Over time, because RAM was the scarcest resource, this use case resulted in hundreds of stranded cores and over 100 TB of stranded storage. Not only is this a huge waste of resources, but it has cost implications to the Service Provider’s managed services, and profitability.
As solutions advance and use cases become larger and more robust, enterprises will want to minimize or eliminate resource drift. The only way to do this is to select architectures that combine the incremental scale of older converged technologies with the simplicity brought by the HCI players.
If you are experiencing HCI drift or you are on the verge of the drift, take a look at Cloudistics and discover how to eliminate resource drift.
Cloudistics has developed a highly composable and scalable private cloud platform, perfect for performance demanding applications and general workloads alike. The architecture combines the incremental scale of converged with the simplicity of HCI with the user experience only seen in the public cloud—with no resource drift.
Cloudistics delivers a complete end-to-end high performance private cloud infrastructure platform that can scale each hardware layer independently. The resources needed to run an application are “composed” on the fly, and intelligent workload placement with SLA enforcement ensures guaranteed performance.
Cloudistics delivers a public cloud experience on a plug-and-play infrastructure solution. It provides the agility, scalability and ease-of-use of the public cloud, but has significantly better performance (8x) and much lower cost (1/2). It comes with all the software and hardware needed to run applications out of the box, and is capable of running all your enterprise applications if you so choose.
To support fast time to value, low cost and ease-of-use, no additional 3rd party software or hardware is needed to run applications. The platform is fully software-defined and virtualized, ensuring agility. Included with the platform is an application marketplace (like an iPhone App Store for enterprise apps) that allows simple point-and-click deployment of applications.
With Cloudistics, what happens if you run out of storage? Simply hot add an additional storage shelf to the system. The intelligent cloud controller will recognize new storage has been added and ask whether it should be added to a specific application or assigned to the pool of resources.
Need more compute? Or additional memory? Or more virtualized networking? The same applies.
With Cloudistics, if your application needs more strawberry ice cream, just add strawberry ice cream and skip the vanilla and chocolate ice cream you don’t need or want. Save your IT budget and profitability by avoiding resource drift.