Orchestrating a Resource Symphony

Orchestrating a Resource Symphony

OpenFlex Composable Infrastructure

Designing, building and maintaining data centers is unquestionably an expensive proposition.

According to a December 2017 Forbes article, “U.S. data centers use more than 90 billion Kilowatt-hours of electricity a year, requiring roughly 34 giant 500-megawatt coal-powered plants. Global data centers used roughly 416 terawatts (4.16 x 1014 watts) or about 3% of the total electricity) last year, nearly 40% more than the entire United Kingdom.”

As the article further details, “This is a big problem, and it’s getting bigger.”

Annual global IP traffic will reach 3.3. zettabytes by 2021. In 2016, global IP traffic was 1.2 ZB per year or 96 exabytes (one billion gigabytes) per month. The traffic will increase nearly threefold over the next five years, and will have an increased 127-fold from 2005 to 2021. (Source: DataCenter Knowledge.)

Adding to the conundrum is the variety of data distinguishing today’s supercomputer landscape; notably unstructured data that includes everything from email messages, videos, and audio recordings to books and social media posts.

OpenFlex F3000

Cognizant of this growing burden shouldered by data center operators, Western Digital (WD) introduced the new OpenFlex storage architecture in early August at the Flash Memory Summit. The move signals an effort by WD to help customers transition from the traditional hyper-converged data center environment to “composable data centers” where disaggregated pools of compute, network and storage are re-deployed or shared in a more efficient manner. In fact, this disaggregation approach demonstrably improves upon both efficiency and total cost of ownership (TCO).

At its most fundamental, OpenFlex attacks the notion of efficiency-obliterating “stranded storage” head on.

To this point, IDC’s Ashish Nadkarni, Group VP of Infrastructure Systems, Platforms and Technologies said, “Western Digital’s OpenFlex architecture provides flexible composability that maximizes resource utilization in a highly scalable and flexible way, as well as eliminates the problems of ‘stranded storage.’ Likewise, its openness eliminates the ‘locked in’ barrier faced by proprietary infrastructures and paves the way for broader ecosystem engagement.”

According to some estimates, hyperscale data centers post approximately 45% resource utilization numbers, based upon utilization over a 24-hour period. Composable infrastructure, on the other hand, is estimated to boost utilization to 70% or even higher. In its own testing, WD asserts that composable infrastructure delivers 40% lower TCO and 50% less initial cost when compared to the prevalent current hyperscale architecture paradigm.

With the advent of OpenFlex, data center operators will be able to independently scale compute, storage and network resources by virtue of software composable infrastructure (SCI). What this ultimately means is that the operator will have a level of resource orchestration capability never enjoyed before.

On the left-hand side of the graphic below, you will see the fixed building blocks that typically comprise a traditional hyper-converged or hyperscale infrastructure. While each of the blocks are ostensibly equipped with ample compute, network, Flash and HDD horsepower, the reality is that much of the resource firepower is rendered moot in the course of actual operation. The fixed architecture that defines the respective building block impedes the operator from deriving full benefit of the resources. OpenFlex untethers the resources so that each of them can be used at full strength (or close to it) as was originally intended.

OpenFlex TCO

In essence, SCI restores the independent status of the resource sets, enabling them to be accessed, allocated and scaled with precision.

The enabling technology foundation that underlies OpenFlex is NVMe™-over-Fabric (also denoted NVMf, NVMef, NVMeoF or NVMeoF™). NVMf delivers the lowest end-to-end latency, while providing high performance), agility and outstanding flexibility.

“We want to break apart the constraints of hyper-converged so that compute and storage resources are just pools that can be composed in a dynamic way, as the workflow dictates,” said Scott Hamilton, senior director of product management, Western Digital.

Western Digital has long been an Advanced HPC Elite Partner. To find out how we can transform your data center – and TCO – with OpenFlex, talk to us today.