5 Comments

Summary:

There was lots of hubbub Tuesday at the OpenStack Summit. Actual customers talking about real implementations of the open-source cloud. Here’s what I learned.

OpenStack Summit
photo: Aaron Hockley

There was a lot going on at the OpenStack Summit, where a couple thousand of the open-source cloud faithful gathered this week. Here are my main takeaways.

1: Customer-to-vendor ratio is getting better.

Mark Muehl, Comcast SVP of product engineering at OpenStack Summit.

Mark Muehl, Comcast SVP of product engineering at OpenStack Summit.

But just a bit. Tuesday’s keynotes featured real, live OpenStack users Bloomberg, Comcast, HubSpot and Best Buy. Not bad.

Bloomberg’s Pravir Chandra said his team set some high goals for what they were trying to  build — they needed high availability, no cascading failures and smooth scale down and scale up. They were able to get there by deploying OpenStack along with considerable custom work of their own, both above and below that layer. They ended up setting up the high-availability databases and figuring out how to aggregate logs from the hypervisor level, said Chandra who heads the security architecture team in Bloomberg’s CTO office

Comcast SVP Mark Muehl said the cable giant is using OpenStack to provide real-time programming guides and fast program search. One application quickly locates the NCAA tournament basketball game you want — no easy task — and brings it up for a fast score check. “We are integrating real-time sports feeds. That app would have been impossible to do on our own [older] platform,” Muehl said.

2: Ceph is hot, hot, hot

Based on an informal poll of speakers and attendees, Ceph storage is where it’s at. The Swift storage system?  Not so hot. Best Buy moved from Gluster to Ceph because of the latter’s self-healing capabilities. Ceph offers object and block storage all in one integrated product while Swift handles object storage only. Mirantis EVP Boris Renski said Swift, which comes out of Rackspace, has lots of production installs, but Ceph is viewed as having a more “elegant” architecture. “Unlike Swift, you can use Ceph as the backend for both object and block.” Also, because of a better algorithm for handling data replication, it can promise better scaling, he said, although Mirantis has not fully tested that out yet.

3: Grizzly brings more maturity and features

Grizzly, the seventh release of OpenStack in three years, brings more features and functions to the table. HubSpot will use Grizzly (with some of its own tweaks) to run images on “full bare metal,” said CIO Jim O’Neill. “That means the same image can run on your cloud of choice … . The application doesn’t need to know or care where it runs anymore.”

And that cloud agnosticism brings huge payback. “We took this single image, picked it up from public cloud into a Rackspace-powered private cloud and saw a 4X increased efficiency running that workload,” he said.

4: Choice is good

The same meme of large companies opting to deploy their workloads on multiple clouds vs. one cloud continued at the show. And yes — I can sense the eye rolls coming — there remains an uneasiness over cloud lock-in. The Best Buy guys put a local traffic manager in front of multiple (unnamed) clouds specifically because “we don’t want to lock into any one vendor,” said Steve Eastham, director of web architecture for the Minneapolis-based retailer.

And even the speakers at this OpenStack event said they will remain flexible in their technology choices going forward if OpenStack doesn’t meet their needs. Asked why Samsung opted to go with OpenStack over CloudStack two years ago, Kirk Kim, cloud CTO for Samsung SDS, said they thought OpenStack would scale better.  “But,” he added, “given the situation today, we might look at that again.” Hmmm. He could not be reached for follow-up.

5: The subtext: AWS and VMware 

The OpenStack faithful are obviously proud of what they’ve accomplished over the past three to four years. It’s not nothing that all that hundreds have contributed to this project and that some customers — outside the OpenStack community itself — are starting to put this stuff into production.

What was left  unsaid, for the most part,  is that OpenStack continues to be measured against Amazon Web Services in the public cloud infrastructure sector and VMware in the (mostly) private cloud market, where legacy applications are in play. This despite the fact that VMware is now an OpenStack Foundation member.

OpenStack may be growing, but it does not have the field to itself.

Feature photo courtesy of hockleyphoto.com.

  1. Thanks for sharing the notes. It would be useful, if you could post on a daily basis. I love open innovation. As some one involved in building various web infrastructure products for more than a decade including Sun Cloud, the biggest enemy of Open Stack is “complexity”. I am currently dealing with complexity of Swift :-)

    Share
  2. It’s an useful article for me. There are 2 lessons that i haven’t known before. Good job!

    Share
  3. It is great to see how OpenStack is coming together so well. I am a huge fan of open source software. It makes the entire IT industry better.

    Share
  4. Barb – great recap of the OpenStack summit. Couple of additional thoughts…

    1) Besides AWS and VMware, other tools I didn’t hear any mention of are products from IBM, CA, BMC and others, some of which interoperate with OpenStack and other cloud platforms. And then there is Apache CloudStack which already has sizable commercial implementations and more to come (e.g., the recent Verizon Terremark announcement).

    2) I hear a lot of talk about OpenStack based hybrid clouds, but I don’t see any tools in the OpenStack suite to enable cross-cloud orchestration. IBM, Red Hat, CA, BMC, and others offer such orchestration tools. But I’d rather see the OpenStack community develop tools that manage cross-cloud movement of workloads and provide governance, metering, and chargeback capabilities.

    Share
  5. The bit about Best Buy moving from GlusterFS to Ceph because of self-healing is pure FUD. It’s old news, based on an old version, and GlusterFS had self-healing even then. Since then it has become both faster and more reliable. Perhaps the Ceph folks feel that their self-healing is better, and maybe they’re even right (I’m a GlusterFS developer so I’m very aware of everything we need to improve), but presenting it as a unique feature is just wrong.

    Share

Comments have been disabled for this post