Blog Post

Mirantis will “solve” OpenStack upgrade woes in the next 6 months, CEO predicts

OpenStack firm Mirantis, which took in a whopping $100 million in funding a couple weeks back, reckons it can fix one of the biggest hindrances to adoption of the open source infrastructure-as-a-service system: upgrades.

Here at the OpenStack Summit in Paris, it’s clear from the record attendance that interest in OpenStack is flourishing, but there are still precious few marquee users showing off pervasive production deployments. In a panel discussion, [company]Ericsson[/company] architecture VP Mats Karlsson complained: “A lot of operators are kind of stuck on Havana because they can’t do the upgrade. When we start installing telecoms workloads these are services with five nines. We can’t afford to take down the system.” Havana is the eighth release of the OpenStack software.

[company]Mirantis[/company], the OpenStack systems integrator that last year introduced its own OpenStack distribution, is apparently on the case. “We’re working on it,” CEO Adrian Ionel (pictured above) told me. “We have a big team working on solving the in-place upgrade problem. We’re talking maybe six months.”

Ionel pointed out that Mirantis has already made its Fuel control plane upgradable using Docker — though Docker may not be the long-term solution for wider upgradability. What’s more, open source being open source and Mirantis not being a vendor of other parts of the stack, Ionel said the fruits of this work would “all go back upstream” to benefit the wider OpenStack project.

“This is part of the big Mirantis value proposition — when people select OpenStack, they want this thin open layer that integrates their entire data center,” he said. “They don’t want an integrated stack. That’s exactly what we’re doing. Everything that we create, we will contribute back upstream, including how to solve the upgrade problem.”

openstacklogoThis isn’t just altruism. Mirantis needs OpenStack to go big if it is to achieve its dreams of an IPO in 2016. “OpenStack needs to become the Android of the cloud, and within that ecosystem we want to be the leading distribution,” Ionel said.

The issue of upgrades is of course closely tied in with that of OpenStack’s release cadence. Right now the system is gaining features at a rapid pace, which manifests itself in six-monthly updates that some users are finding a bit hard to handle.

During the morning keynotes, BMW data center chief Stefan Lenz was full of praise for OpenStack’s promise of an industry standard for IaaS deployment, and for its open source nature, but he said the German carmaker wasn’t putting high production workloads on it just yet.

“We are conservative people,” he said. “The thing that bothers us most is the release cycles and the huge changes we have from release to release. We need more stability still in the future.” (To be fair, he also identified a major problem in BMW’s need to adapt its existing processes around OpenStack’s self-service ideal.)

OpenStack Foundation COO Mark Collier said in a subsequent Q&A that, as it’s open source, “everything’s on the table”, including the six-month release cadence. However, he added: “One of the key reasons why the speed has been frustrating is because the upgrades are harder. As we’re making the upgrades easier…the fact that you’ll want to be upgrading more often won’t be as much [of a problem].”

“Just because there’s a community release every ix months, doesn’t mean someone using commercial implementations will have to upgrade on that same cycle, OpenStack Foundation chief Jonathan Bryce noted.

Ionel’s take? Six months is a good cadence, but “OpenStack has spent an enormous amount of time on the plumbing and it needs to spend a lot more time on the end users and the developers… Stability and user experience are not so much a matter of cadence as [they are] of priorities. It’s a matter of where you put the emphasis.”

One Response to “Mirantis will “solve” OpenStack upgrade woes in the next 6 months, CEO predicts”

  1. I understand how this would be helpful for organizations choosing OpenStack for new projects six months from now. But, organizations that are on Havana, Icehouse or even Juno today have a big challenge. OpenStack has similar problems to ERP system upgrades of the past. Once you add the customization that you need to add-value then you experience the upgrade pains. After years of practice, it’s still no walk in the park for ERP vendors

    I see the challenge as moving from a highly customized Havana installation to an installation that provides a better future upgrade path. There’s unavoidable pain in making the transition.

    How many upgrade cycles will it take before a telecom provider sees the value in making the transition from their current platform? I’m skeptical. If you are on a homegrown distribution today, you may end up forking installations and migrating workloads opposed to upgrading an existing installation.