Subscribe
About

You've been served

Buying a server is easier than ever. But which route to go?

Paul Furber
By Paul Furber, ITWeb contributor
Johannesburg, 13 Jul 2015
JJ Milner, Global Micro Solutions
JJ Milner, Global Micro Solutions

It's never been easier or cheaper to purchase server capacity, whether cloud, virtualised or bare metal. Amazon, Google and Microsoft, the three biggest public cloud service providers, are in the middle of a price war to see who can go lowest. The dedicated server providers are also offering very competitive deals on their server instances, whether bare metal or VPS, and there are a lot of them: Rackspace, Softlayer, Hetzner, Linode and Firehost, to name but a few.

At the same time, global datacentre spending is almost flat. Gartner's Worldwide IT Spend Forecast for 2015 says it's projected to reach $142 billion in 2015, which is only an increase of 0.4 percent from 2014. Of all datacentre components, it says servers are under the most pressure. "The server segment has seen the biggest impact due to the greater pricing pressure that server vendors are exposed to, due to their relatively lower margins," says the company.

It's clear that there is considerable commoditisation pressure on the server market as a whole, a trend that was kicked off by virtualisation and then boosted by widespread cloud adoption. Virtualisation in South African companies is now very widespread, says Clive Donninger, CIO of Associated Motor Holdings.

"I think every man and his dog, even in the small organisations, has already gone virtual. The only reason to go physical is if you have an application that cannot be virtualised.

"I think in the beginning, many people were really cautious - fear of the unknown. But if you're going from physical to virtual, Microsoft and VMware both have sophisticated tools to assist with sizing.

"The vendors will run these tools for you free of charge and come up with accurate sizing. Once you've virtualised, you order additional capacity as you go along. DR is also much easier. If you have two virtual clusters in separate server rooms, DR virtually becomes automatic. Thus, for most people to virtualise is a no-brainer."

An essential niche

Donninger's next decision, when he has to replace hardware, is whether to purchase and own or rent in the cloud. "The answer will come down to a detailed costing exercise. Will renting be cheeper? The devil is in the detail," he says. "The other factor is whether the business wants to go capex or off balance sheet."

The devices that run your encryption, you don't want embedded in someone else's datacentre.

Simon Hudson, Amalgam

JJ Milner, founder and chief cloud architect for Global Micro Solutions, says his company buys hardware both because it's a cloud provider and because there are some applications that aren't fast enough or certified to run in a virtualised environment. "We certainly buy infrastructure to offer to our customers. But certain things won't run in a virtual environment. Others don't have enough throughput so you do need physical tin for some roles. And also there are some applications where the vendor hasn't gone down that road."

Is the server itself interesting anymore? Milner says not really.

"It's the things around the server that are interesting - power, cooling, skills, licensing, and flexibility are really what drive things today," he says. "The hardware is the least interesting part of it." Simon Hudson, CTO at Amalgam, agrees.

"The cooling, licensing and flexibility are the most interesting parts for me. A move to a hardware virtualisation platform is a move to get away from the responsibility of all that hardware-related stuff: the warranty, hardware, licensing, cooling and electricity. It does depend on your business model and how much pain you're willing to accept if you devolve certain facilities and don't get it right."

Hudson says there's another essential niche left for hardware that you own.

"The devices that run your encryption, you don't want embedded in someone else's datacentre."

For certain companies, such as banks, those tend to be large, heavy devices that live in their own datacentres, preferably underneath a bomb-proof shelter.

The converged alternative

But for some companies, managing any of their own infrastructure piecemeal, no matter how virtualised and efficient, is becoming too slow and too inflexible. For them, the answer is to invest in an all-in-one solution in which the servers, the networking and the storage are included: converged infrastructure.

It's a market that's grown rapidly since the end of 2009 when Cisco, EMC and VMware announced a joint partnership called VCE to offer infrastructure packages called Vblocks. Instead of ordering, installing and configuring all their servers, networks and storage, Vblock customers get a blade chassis, a tiered SAN, a complete virtual environment and network switches with a GUI to manage everything. This can be deployed in a couple of racks in less than a month, and provisioning hosts, VLANS and data stores takes hours. The growth in Vblock sales has been so large that it's propelled Cisco, traditionally a networking only vendor to number two in global sales of blades, behind HP, which also has a converged infrastructure offering of its own.

If you're adopting converged infrastructure, your organisation has to reflect that.

Bevan Lock, Lenovo SA

One of the downsides to all this convenience, says Bevan Lock, currently technical sales lead at Lenovo SA and former technical sales lead at IBM SA, is that it can be difficult to integrate into an existing environment.

"Converged infrastructure is easiest when you're going into a green fields environment and you don't have existing servers, existing hardware or existing anything else. We've had success selling it to smaller organisations or for more specific projects."

Lock says the other impact to an organisation investing in Vblocks or any other kind of converged infrastructure is on the staff. "If you're adopting converged infrastructure, your organisation has to reflect that. Established IT organisations don't see infrastructure in general as converged, they see it as silos. They have guys who understand networking, a storage team, a network team, a compute team and some guys who understand virtualisation. To invest in converged infrastructure, you have to overcome the mindset in the organisational structure that says why should the networking guys want to give up their territory to the storage or theserver guys?"

Some larger companies may not have a choice. If remaining competitive is based on speed to market or ability to innovate, then an all-in-one solution may be the only way to deliver on that. But there are downsides, not the least of which is a divergence in priorities of the vendors.

"When VCE came along, it reminded me of Itanium," says Lock. "Itanium presupposed that you could keep Intel, IBM, HP and SCO aligned. While at a point in time your strategies might align, over time they diverge."

And that's exactly what has been happening to VCE: VMware has been getting into networking by acquiring Nicira, EMC bought Cloudscaling and Cisco slashed its stake in the partnership last October down from 35 percent to ten percent, as well as getting closer to EMC rival NetApp. For the server buyer, keeping an eye on vendor strategy is still a vital part of the strategic planning.

This article was first published in Brainstorm magazine. Click here to read the complete article at the Brainstorm website.

Share