The cloud has been helping lower the cost and improve the quality of technology infrastructures and business practices for more than a decade, but implementation of cloud technologies can still pose challenges.
That applies no matter whether you are an enterprise implementing your own private cloud, a service provider constructing the purpose-built infrastructure it needs to serve customers, or somewhere in between trying to weave both worlds together into hybrid cloud services.
And hybrid cloud is really where the action is: an NTT/451 report here found 84 per cent of enterprises are implementing hybrid cloud, with 63 per cent working on a formal strategy. Only 16 per cent plan to employ just a single cloud.
With the global cloud services market showing strong growth and predicted to hit $300bn by 2021, there is plenty of opportunity for enterprising vendors willing to step in with the kind of technical assistance that helps cloud service providers (CSPs) and enterprises make the right hardware and software choices for building the increasingly virtual, integrated and distributed clouds.
Listening to the hype emanating from certain sections of the industry, you’d think cloud was a simple case of install and go, integrate and run, virtualise and fly. Some vendors certainly like to give the impression that all you need do is install their servers and/or storage and deploy their software and you’re done. But the world of integrated systems, services and elastic compute is far more complicated. It’s not insurmountable, just complicated.
For example, the hardware must be capable of handling anticipated workloads in addition to those unexpected – to be able to scale in line with expectations and handle both planned and unplanned peaks. Workloads will place exacting and specific requirements on both network and storage. Adding further complication is the sheer number of vendors to select, vendors whose products and services increasingly make up the hybrid-cloud mix. And what of interoperability? It’s the bread and butter of hybrid cloud, but you should expect complications getting products and services from one supplier to work with those of another.
What this demands is a good deal of testing and evaluation.
Enter Supermicro and Intel with a Cloud Center of Excellence (CCoE) at the former’s Netherlands’ site in Hertogenbosch, offering a facility to run proof-of-concept testing or evaluate the performance of Supermicro and Intel configurations, both remotely or directly on site.
The target market is primarily what Supermicro and Intel call the “next wave” of CSPs. These, according to Supermicro’s senior vice president and chief product officer Raju Penumatcha, are new and emerging service providers delivering speciality or regional cloud services to meet more specific needs than cloud giants such as Amazon with AWS and Microsoft with Azure. The CCoE will also help large enterprises that want to build their own private or hybrid clouds.
The lab’s test-bed equipment includes 25 Gb/sec networking with secure remote access via OpenVPN to systems that feature some of the latest advances in compute, storage and networking technologies from Intel. CSPs are likely to be interested in testing out Supermicro’s BigTwin servers, which offer up to four individual nodes into a 2U enclosure or the Ultra family of high-density 1U and 2U servers, while the 1U SuperStorage system can be configured with up to 32 of Intel’s “Ruler” form factor SSDs for a full petabyte of NVMe storage.
The CCoE is being offered for two common areas that can help speed up delivery of cloud services and minimise the cost of pilot projects. One of these is a proof-of-concept, where Supermicro builds a configuration for customers to deploy the kinds of workloads they intend to run and thereby test how those would perform. The other is for demonstrations, where customers can try a specific platform, such as OpenStack, to gain experience and evaluate how it may meet requirements, without having to incur the cost and risk of first procuring and building their own infrastructure.
The facility has been running in stealth mode since August with a number of customers already having taken the plunge. According to Penumatcha, uptake has already exceeded the expectations of both Supermicro and Intel with the number of proof-of-concepts already numbering in double digits and the facility generating revenue in double-digit millions of dollars.
A pattern is emerging in what types of systems are being evaluated.
Penumatcha says many of the proof-of-concept trials have involved customers evaluating VMware’s VSAN or virtual desktop infrastructure (VDI) deployments, while in other cases they have asked to bring in ISVs that work with archival or data management applications.
Supermicro, meanwhile, is already looking at expanding the scope of the CCoE. One area that has been an important market for the company is HPC infrastructure and Penumatcha indicated the firm is likely to open the lab facility for testing HPC configurations.
“Many major service players are adding HPC services to their cloud offerings, which are incremental to the traditional HPC build out. In this context, given our priority for the CCoE are CSPs and large enterprises, we would support testing HPC workloads, within the limitations of infrastructure, as a proof-of-concept,” Penumatcha says.
Intel’s view, however, is that there is an opportunity for the CCoE to serve customers in another area: in keeping abreast of, and getting a taste of, new and emerging technologies.
Norberto Carrascal, Intel’s EMEA business consumption director, told us: “Technology is evolving rapidly, especially within this environment. Just to keep up is not enough. You have to continuously understand the new products, the new offerings, the new technologies, the new features and the new capabilities, because you need to bring them to your portfolio if you want to be successful at selling to end customers.”
Carrascal reckons their joint lab is the “perfect” environment for cloud service providers to conduct tests. He lists some of the options available through the facility. “When you think of some of the latest products, they [Supermicro] already have early access to next-generation Intel Xeon Scalable processors – codenamed Cascade Lake – and Intel Optane DC persistent memory; they have Quick Assist Technology that is going to allow customers to do faster encryption; and they have the innovative Intel Optane DC SSDs,” he says.
Cascade Lake, of course, are Intel’s planned next-generation of Intel Xeon Scalable processors for servers, due in the first half of 2019, while Intel Optane DC persistent memory is a new class of memory to be delivered with the new processors that will enable non-volatile storage-class memory to be fitted in DIMM slots and be treated as a new tier in the memory hierarchy between DRAM and flash storage.
Supermicro, meanwhile, is also a partner in the Intel Select Solutions program, which serves specific areas of customer interest such as advanced analytics, hybrid cloud, storage and networking with rigorously benchmark-tested and verified systems optimised for real-world performance.
A side effect of the CCoE of course is that customers running a proof-of-concept on Supermicro systems choose the same hardware when it comes to rolling out their production infrastructure.
This is something Penumatcha warmly anticipates: “The objective is to grow our direct sales and go beyond the component or building-block type of business model we have traditionally been known for, and be a little more centric around Intel technologies. It’s something that we wanted to do jointly, with leads coming from both Intel and Supermicro,” he said.
One thing is clear: there exists a ready market for CCoE. With the increasing complexity of IT infrastructure and the potential costs associated with making the wrong infrastructure choices, a facility that helps IT evaluate new capabilities, road test new systems and experiment with combined stacks but without shouldering the risky upfront investment will prove invaluable.