kentoh - Fotolia
The rise of the ODMs: Should the branded hardware suppliers be worried?
The use of ODM datacentre hardware makes good economic sense for the hyperscale providers, but what about everyday enterprises?
The big computer hardware suppliers have ruled the hardware roost for many a year. So assured was their hold on enterprise IT budgets, there was even the saying of “no one ever got fired for buying IBM”, alluding to how these big IT brands carried the argument for large hardware expenditure. But is this now at risk?
Public cloud adoption is rising, and anyone who’s had the pleasure of wandering around some of the datacentres used to host these clouds may be surprised by the lack of large-brand logos on the hardware inside, and in many cases, the lack of any logos at all.
Many of the largest public cloud providers have eschewed the big manufacturers in favour of using cheaper suppliers who can build the datacentre hardware they need to their precise specifications, and for a better price.
This is where the original design manufacturers (ODMs) come in. The big public cloud providers can go to them with a specific design and demand 100,000 or so servers over the next 12 months, and the ODM can build the boxes using commodity componentry in countries with low labour costs.
The end result is tailor-made hardware priced to suit the cloud provider, while still providing a good enough margin to the ODM.
The largest public cloud provider, Amazon Web Services (AWS), carries out a series of tests to calculate how much failure it can deal with in its datacentres.
As part of this, its ODMs are asked to build to a defined failure level – one where costs can be cut while still remaining within AWS’ own tolerances. Branded hardware suppliers, however, are loath to build hardware of a lesser standard than its normal commercial output.
Read more about ODM adoption trends
- IDC’s latest EMEA-wide server market sales tracker suggests soaring demand for ODM servers is hitting tier-one manufacturers hard.
- Traditional server providers still have the lion’s share of the on-premise datacentre market, but original design manufacturers are now a major force.
The reason being, should this hardware make its way on to the grey or second-hand market, it would reflect badly on the brand. Yet for organisations that fully understand how a highly-virtualised platform can deal with failure, it is a good option – and it may be a market where the brands are missing a trick or two at the moment.
ODMs have other economic benefits working in their favour too. They don’t sell through any channels, do not have large sales forces, and do little to no marketing, making it easier for them to run lean and mean.
That said, ODMs may not offer much in the way of support, either. The buyer is expected to become the full owner, not only of the hardware, but its full operation. The big brand hardware suppliers cannot do this as they have so much extra overhead to carry.
AWS has been a leader in this approach, buying more servers per year than any other provider, and it can exert a lot of pressure in the market. Sure, Dell, IBM or HP would love to be one of its prime hardware suppliers, but it just doesn’t work for AWS.
AWS wants a platform that is essentially designed by itself and needs suppliers that can build to that specification – and can rapidly build in changes as and when required.
Imagine IBM, with its multiple research centres, its massive patent library and its multiple tiers of account management just tugging its forelock and saying “whatever you say, my liege” to AWS. It is unlikely to happen, and the whole process would get bogged down with a group of distinguished fellows trying to persuade AWS that the use of Watson-as-a-service running via Quantum Computing on SoftLayer would be a better way of doing things.
Taking ownership like AWS
By taking responsibility for everything from the motherboard to the facility, AWS is in total control of its own future, and it credits this decision with giving it a solid, competitive advantage. Its ODMs do what AWS says, and both sides are happy with that arrangement.
Facebook takes the same approach, and in the throes of updating the hardware in its datacentres using its own server designs.
The company has open sourced a large part of its design under the Open Compute Project (OCP), so others can create a similar platform should they so desire. Even Microsoft, a long-time Dell partner within Azure Cloud, is moving to an OCP hardware design under its Project Olympus strategy.
And here lies the crux of the matter in the battle between ODMs and branded manufacturers.
Few, if any, organisations are large enough (or have the same infrastructure needs as an AWS or a Facebook) to go to an ODM and sign a deal that makes economic sense to both sides.
They may be able to take the OCP design and purchase a standard unbranded ODM product, but that is just a general low-cost supplier-customer relationship with little capability of adding competitive advantage through differentiating the platform itself.
For the average organisation, buying hardware from a known entity with full commercial support still makes sense. However, as time moves on and more workloads move to the public cloud, this is a decreasing market for the main hardware suppliers. This is why IBM has spent so much time and energy repositioning itself as a purveyor of servers, network equipment and storage to a “cloud-first” company selling capabilities around SoftLayer, BlueMix and Watson.
Is all lost for the other hardware suppliers, such as Dell and HPE? Not necessarily. Outside of the massive public cloud providers, there are many other smaller public cloud providers only buying a few thousand servers. Although a basic ODM approach may be tempting, there are still opportunities here for the branded suppliers to offer a fully supported stack.
In football parlance: they think it’s all ODM... Well, it isn’t yet.