Page 13 - FCW, June 15, 2016

HCI’s inherent scalability facilitates starting small and rapidly growing, making
it a low-risk alternative to traditional large, monolithic systems and obviating the need for lengthy, error-prone planning exercises.
foundation, with products differentiated by their flash performance and sophisticated use of storage services such as inline compression, deduplication, encryption, real-time replication and archiving.
A niche that’s particularly well suited to HCI is the capture and storage of video surveillance data. Another scenario that’s becoming popular involves using HCI to aggregate Internet of Things sensor data. Linear scalability of capacity with increasing node count is the reason HCI works so well for video capture.
• VDI. Virtual desktop infrastructure is a great HCI appli- cation because of its inherent scalability. By identifying the resource needs of the typical VDI user — for example, required CPU, memory and storage — organizations can build an HCI system with a known performance profile and user capacity. As the user population grows, organiza- tions merely add HCI nodes, thereby scaling performance in line with capacity.
VDI is the reason the U.S. Army turned to HCI when it sought to replace 40,000 physical desktops. VDI has a well- earned reputation for being slow, a problem the Army saw in early pilot testing. After some testing and despite having invested seven figures in a legacy VDI system, the Army’s chief virtualization architect was convinced HCI was the better design, saying, “The performance was off the charts compared to the previous solution.... It was definitely a physical desktop replacement solution, and they loved it.”
The Army’s deployment is hardly the government’s only VDI success story. As FCW’s sister publication GCN report- ed last year, Sandia National Laboratories used a similar design for its 1,500-user (and growing) VDI deployment. • General-purpose workload consolidation. The heart of HCI systems is a virtualization stack that often but
not always uses VMware vSphere. Indeed, the software stack for HCI storage appliances runs as a set of VMs. Given the performance of today’s CPUs, HCI nodes work great as general-purpose virtual servers. And given the high-density of HCI designs — which often have 20 to 40 cores, 256G RAM per node and four nodes per chassis — HCI makes a great platform for workload consolidation. By managing all nodes from a single console, whether vCenter or the vendor’s own tool, HCI facilitates putting more and more workloads on a single cluster without increasing IT overhead.
The Office of the Secretary of Defense consolidated and migrated more than 2,000 servers — including all types of information, data and workloads — to an HCI platform. The net results were an 80-plus percent reduc- tion of the office’s data center footprint and the replace- ment of 60 racks of existing equipment with just 10 racks of HCI systems.
• Remote office/branch office infrastructure. A corol- lary to the consolidation use case is HCI as an all-in-one ROBO platform. The combination of the ability to run arbitrary virtual workloads, local storage that includes enterprise services such as replication and deduplication, and multiple network interfaces means a single HCI box can serve even medium-sized facilities. Indeed, by host- ing virtual network appliances such as wide-area net- work accelerators, virtual private network gateways and intrusion-detection/intrusion-prevention software, an HCI system can replace several stand-alone network devices.
According to Pivot3 Federal CTO Eric Oberhofer, agen- cies are already starting to use ruggedized HCI platforms in field deployments on aircraft, ships, trucks and other mobile scenarios.
Agency action plan
HCI is well suited for many agencies as they try to serve more users and applications with the same budget. IT leaders can consider HCI for specific needs such as VDI and large-scale data collection and as a platform for vir- tual workload consolidation.
But regardless of the scenario, HCI’s inherent scalabil- ity facilitates starting small and rapidly growing, making it a low-risk alternative to traditional large, monolithic systems and obviating the need for lengthy, error-prone three- to five-year planning exercises. In the long run, HCI is a great platform for private infrastructure- and platform- as-a-service stacks such as Microsoft Azure Stack, Open- Stack, Cloud Foundry, Red Hat CloudForms and others that are the foundation of next-generation, cloud-native applications. n