Crushing the three tiers of a data centre into a Lego brick appliance. Those are the words John Bainbridge, account manager at EOH Technology Services, with responsibility for the hyper-convergence technology set, says he would use to describe Cisco's Hyperflex solution to his grandmother.
"The three tiers are your server component, your storage component, and your hypervisor sitting above it; add the software-defined network and the solution is complete," Bainbridge says.
Natius van der Watt, product sales specialist for data centre at Cisco, likes Bainbridge's simplified approach, but says technically, Hyperflex is an out-of-the-box solution of servers and switches, with software pre-installed. "It is the cherry on top of a virtual machine solution," he says, explaining that the real value of Hyperflex lies in its management capabilities.
"With some other vendors' solutions, you still have to have a separate and brand new management interfaces to manage their software - one for networks, one for servers, one for utilisation, one for storage. But with Hyperflex, there are only two management points: one for hardware and one for software. All hardware is managed from the UCS Manager, so if you are running UCS Manager already, you have no additional management platform to pay for or integrate. It's just a plugin on a VMware Virtual Centre, just an add-on to manage your Hyperflex storage platform. Similarly, Hyperflex software management is a plugin in the Virtual Centre, the management tool for your hypervisor, where Hyperflex simply becomes one of the tabs," Van der Watt says.
Bainbridge says crunching those three layers into a single Lego block delivers better economies of scale, reduced power consumption, reduced rack space, simplified management and "a lot of soft benefits that come with being able to reduce the footprint".
He says other benefits include:
* Improved performance - "In legacy architecture, the storage isn't situated close to the processing layer, but the closer those two are together, the faster and the more performance you will get for your application. This solution also introduces a solid state drive into your storage subsystem, making it faster," Bainbridge explains.
* Fewer overheads to providing services - "This solution reduces the amount of infrastructure, the number of resources, and, in most instances, the number of management panes required. This allows you to re-purpose people to do more in the virtual and application layer, rather than being traditional storage jockeys," Bainbridge says.
* Single pane of management -"Cisco has taken management a step further and included the whole networking component into the Hyperflex solution - at the moment they are the only vendor that is really doing that. I suspect it comes from their experience of being part of the VCE alliance. If you look at flexpods and vblocks, they were hyper-convergence 1.0, and when you look at the Hyperflex solution, it's very much hyper-convergence as it is intended to be."
True hyper-convergence, he says, is the ability to provide all services below the hypervisor in a simple, logical, cost-effective manner. "You're looking to get the best of both worlds, to get cloud economics, flexibility, speed provisions, but with enterprise-grade requirements of seeing neighbours and protecting information."
* Agility - Traditional, legacy data centres are "unwieldy and their design has become accidental", says Bainbridge. "A lot of things have arrived in the data centre over the years that the original infrastructure wasn't designed to deal with. Virtualisation took a physical environment and made it very difficult for that environment to keep up with a very agile solution. The hyper-convergence in Hyperflex for the first time gives you the ability to be as agile as the hypervisor."
Bainbridge believes customers running Cisco UCS will find Hyperflex an easier transition into a proper hyper-converged infrastructure. "They'll be able to easily get the goodness and value out of their existing purchases."
Van de Watt says there is a nuanced approach needed: "As Cisco, we realise that Hyperflex is a new market, a new space in the market, and that hyper-convergence in general does not have the wide berth of workload that physical infrastructure has. We don't profess to run a massive rack-scale Oracle database as efficiently; we guarantee virtual desktops and non-production environments. We can get a Hyperflex solution up and running in between 60 and 90 minutes. Hyperflex has been designed for the VM environment, but if you are looking at a massive database, you would prefer to go with a more traditional structure."
Share