Dell, Intel and pal pump $10m into OpenStack biz Mirantis
Chip titan and Chinese sugar daddy help bankroll PC giant's cloud dream
Michael Dell’s computer giant, Intel and WestSummit have joined forces to inject a new open-source cloud vendor with $10m so that the startup can develop and sell its delayed OpenStack public cloud.
The investment arms of the three tech giants - Dell Ventures, Intel Capital and China-based WestSummit Capital - awarded the cash to Mirantis, an engineering outfit with 30 OpenStack deployments under its belt. Mirantis customers include Hewlett-Packard, NASA, Gap, PayPal and AT&T.
Mirantis had been working with Dell “on a number" of OpenStack customer engagements with Dell for 18 months so there is an element of self-interest at play for the computer giant and aspiring OpenStack service supplier.
“This was a logical extension of this relationship,” Mirantis co-founder and executive vice-president Boris Renski told The Reg of the funding. “We will have a more formal relationship going ahead – Dell will use us in their channels to sell OpenStack. Dell has [also] announced they are building Dell public cloud on OpenStack – we are the company building the cloud for them.”
Nnamdi Orakwue, vice-president of Dell Cloud, said in a statement: “The partnership with Mirantis demonstrates our commitment to the community and our goal of becoming a leading contributor to OpenStack.”
The $10m joint investment will be spent on expanding R&D, including the development of a library of pre-packaged APIs to deploy and manage OpenStack that Mirantis might release commercially. It’ll also pay for 15 of the company’s 90 engineers to contribute to the OpenStack project – up from four.
Dell had promised an "open source cloud" in 2012 and another one based on Microsoft Azure before the end of the year. Dell’s OpenStack cloud is now expected by the end of 2013, but there’s no word on the coming of Dell’s Azure cloud.
Dell is now behind HP, who launched its Cloud Compute OpenStack service in December.
Mirantis is working with Dell on requirements and specifications for its OpenStack cloud and on architecture and coding. Renski reckoned Dell's deployment is relatively simple because Dell is swallowing what he called the OpenStack “trunk” – compute, storage, block storage and quantum networking.
He said there's no reason from a technology perspective for the rollout to slow down, given that Dell has gone with the "trunk", but he did talk about a few hurdles that the firm was anticipating and dealing with along with some other unexpected challenges.
On the "anticipated hurdles" side, he cited the complexity that often results when the software meets the hardware, as drivers must be written to work with the different metal.
What's going on under the bonnet?
“You get peculiarities that relate to the underlying hardware architecture, such as what hardware are you using for storage? If it's for Dell, then you need to build logic drivers and a deployment topology – single or multiple controllers, controllers with self-healing etc. Those map on to the hardware architecture and have to work,” Renski said. “This is where complexities come.”
This is a big challenge, particularly given that Dell’s strategy on the cloud mirrors its trademark policy on the PC and server: giving customer choice and customisation options.
Dell’s plan is to make OpenStack available on its hardware with support for a variety of components that can be configured according to preference. Users will then host OpenStack on Dell hardware in their data centre and head over to the Dell cloud when there’s a need for additional compute or storage.
The great unexpected, however, will descend once Dell's cloud rolls out and updates to Folsom are needed. Folsom is the latest version of OpenStack; Dell's computer, server and cloud rival HP released Cloud Compute on the older Essex release of OpenStack. HP plans to update to Folsom.
Updates are a big one, because while many issues can be foreseen in the planning phase and during testing in the labs, other unexpected issues can crop up in a live environment. This, Renski reckons, is what gives him the fear.
“Once you launch to the public you want to keep the cloud and sync with the upstream code base and not disrupt the loads running. Few people have had experience with that, including us, and we anticipate some challenges down the road,” Renski said.
“We’ve done this kind of stuff at the lab level, but when you work at scale you never know what’s going to pop out. Unless the methodology to do the upgrade has been done at scale across the industry, you can get things that you aren’t expecting.”
Dell’s cloud will sit on 1,000 nodes at three data centres; it’s being developed on Ubuntu and is expected to support Ubuntu, with CentOS, SuSE, Red Hat and Windows as guests. A thousand nodes might not sound big in terms of Google of Facebook, but Renski reckons the nodes aren’t the immediate goal.
“We’ve done 1,000 node implementations before: it’s a manageable scale. The goal is not get that up to 2,000 nodes, it’s to get it to work at 100 nodes with the proper mix of features and have all the features,” Renski said. ®
Sponsored: DevOps and continuous delivery