Archive for the 'Technology' Category

Next Steps for Openstack; Goodbye Austin, Hello Bexar

So, we’ve kicked Austin,  the first official release of OpenStack, out the door. What now?  If you were hoping for a little rest and relaxation, you joined the wrong project.

In two short weeks we have the OpenStack Design Summit (ODS).  ODS is not a conference, where people lull you to sleep with presentations.  ODS is a working summit.  We have to plan, discuss, and write specifications for the next two releases of OpenStack.

So how does this work, you ask? The entire process is documented on

  • Submit blueprints. It all starts with blueprints.  Blueprints are short descriptions of features, or parts of features.  Landscape has a good description of blueprints, if you want to learn more.
  • The team of Openstack architects then reviews the blueprints and , if needed, asks for more information from the filer.
  • We then choose the blueprints we think are the highest priority and schedule discussions about them at ODS.
  • After the Summit, the full specifications are written, I will approve and prioritize them.
  • As you you start developing, please update the status of your blueprints
  • Don’t forget to link your development branch to the blueprint using the ”Link a related branch” link on the right hand side of the blueprint page.

What People Are Saying About OpenStack and What They Should Be Saying

I am a bit irritated at some of the criticism and, to be honest, some of the praise the OpenStack project has received.

Lets deal with the criticism first:

Vaporware: I have seen the project called vaporware in several forums.  This doesn’t seem to be an appropriate use of the term, and smacks of FUD, but may just be ignorance or unfamiliarity with Launchpad, where we host our code.  Some links to help you get to things.

Nothing is usable: I’ve seen this too, usually followed by a statement implying that it might take years, for it to be production ready.  The OpenStack Objectstore is already in production at Rackspace holding many dozens of petabytes across multiple data centers.  Please try to install it and give us feedback.  Now, the compute side is a slightly different story.  I believe that NASA has a version of this runnig in production, but development is moving so quickly, unless you want to help us develop or test it, I recommend that you wait until our first release in October.  Some examples of the quick development of OpenStack Compute’s first two weeks:

This is just a marketing/PR move: I can say that this is not true for Rackspace.  I can’t know the motivations of everyone involved, but I was a part of the team that convinced Rackspace to do it and none of the technical or business people involved mentioned PR or marketing as a reason to do this. This was not an easy decision and needed to be explained to the organization from top to bottom.  We told them the same thing we are telling everyone else.  “It is the right thing to do”.  Then we made the following guarantee: Rackspace will sell just as much software next year as it sold last year: None.

OpenStack will not kill open core: OpenStack is a project, not an idea or a philosophy.  We have no plans to take on any way of thinking.  We only got involved in this debate because I feel very strongly about it and wanted to make sure that OpenStack’s position was understood.  We are strong supporters of open source and the ideals it stands for. We feel like open core, as it is most commonly implemented, is a bastardization of those ideals, however, OpenStack is not Don Quixote and we don’t feel like charging at windmills.  Our real goal is to level the playing field in the IaaS ecosystem and no longer allow people to act like creating a IaaS fabric is some kind of super secret magic.  Let’s do it once, give it to everyone, and move on to more interesting things.

Now the Praise:

OpenStack will kill open core: Please see above.  Killing or not killing open core is not one of our goals, however, not being open core is one of our goals.

OpenStack takes on Vmware/MS/Amazon: We are not doing this as a tactical move against any specific company, but because it should be done and there is no reason not to do it.  Our mission statement from the wiki:

The OpenStack Open Source Cloud Mission: to produce the ubiquitous Open Source Cloud Computing platform that will meet the needs of public and private clouds regardless of size, by being simple to implement and massively scalable.

This is as big as the beginning of Linux or Apache: We are all excited, but please don’t set us up to disappoint.  I really hope we do something great, and both of these projects inspire me, but let’s remember how long it took for these projects to become the giants that they are.  I don’t want to down play what we are trying to do, but let’s do it first and let the results speak for themselves.

What would I like for people to say?

Well the truth, good and bad and with out hype or malice. Here are some examples:

The bad (well not so bad, but do you blame me?):

OpenStack needs to organize quickly: We are getting a huge response and we need to make sure we can handle the influx of developers and code.  The most important things to me is that we maintain quality and that we deliver when we promise.  If we can do those things we will be successful.

OpenStack needs to  publish it’s governance: We do.  This is valid criticism.  We are working on it based on feedback from the community. Once it is done, it needs to be approved by quite a few parties.

OpenStack sucks because it should have feature a,b, and c: Should we?  Well write it if you can or pop into #openstack on freenode and ask us to write it.

The good (or what we would like to hear):

Wow this is great code: Thanks.  We do the best we can.

We really like the open way you are running OpenStack: Thanks, we feel that to be successful we need to have an open collaborative process.

I want to help: We would love your help.  Can’t code?  Come help with community or write documentation.

Next Stop UDS

All the Canonical folks are meeting this week prior to UDS.  We’ve already had some great discussions including:

Canonical Plan for World Domination

— text redacted —

(Sorry, if I told you, I’d have to kill you.)

Actually, it is a great time to see all our co-workers from across the globe.  However, I can’t wait for next week, when UDS starts.  UDS is a great time to see all of you, the people that make Ubuntu great.  Please come up to me and talk to me, even if I seem busy.  We are randomizing the tracks this UDS, so each track will move to a new room after each session.  There is no more hiding in a room and not meeting new people.  Please take advantage of it.  The F/OSS world is a weird diverse and interesting place, don’t let yourself miss any of it.

UDS Barcelona Has Incredible Potential

This UDS each track will have two session rooms, plus a breakout room. That means we will have, at least, twice as many sessions. We should some out of it with twice as many specs for new features. My question is, how will we handle twice the data?

In previous UDS’ the server track was very well documented, thanks to Adam Sommer from the documentation team.  Adam was able to attend all sessions, and the documentation was uniform, because one person was responsible for the format. I think that because of Adam’s excellent work, we had some of the best documented sessions at UDS.  This was good, because I am usually suffering from information overload by the end, and unable to remember my name, let alone the critical details of the discussions.  We will have to come up with another solution in Barcelona.  Adam will only be in some of the sessions, so the rest of us will have to pick up the slack.

If we take good uniform notes, we could come out of this UDS with more data from more discussions than ever before. This will ultimately result in a better Ubuntu.  All we have to do process all the data in a usable way.  So, you will probably hear me start every session with a reminder about note taking and a quick review of the format we want them in.

Free as in beer:
As usual, Server Team community members find me at UDS and I will buy you a beer.

I’ll be at the Free Software & Technology Expo in St Louis, MO on Saturday

I’ll be speaking at the Free Software & Technology Expo tomorrow May 9th.
My talk will cover:
* What’s new in Jaunty, including the Ubuntu Enterprise cloud
* Hints of what might come in Karmic, and a UDS preview
* Observations from the front line (or What it’s like to work for Canonical, and the furious pace of our 6 month release cycle

If you are in or around St Louis, stop by and say hello. I’d love to talk to you!

Exchange Proxy Anyone?

At UDS, there was talk (mostly from Dan Shearer) about using Openchange to provide a mapi proxy to MS Exchange.  I find this interesting.  If Dovecot/Postfix could use it, it would allow *NIX clients to talk to Exchange without turning on OWA.  This would delight sys admins on both sides of the fence.  Of course, Outlook clients could also talk straight mapi to the proxy as well.

I like the idea of abstracting MS Exchange, because once we have all the clients talking to us instead of MS Exchange, it makes it much easier to replace exchange.

The Openchange folks have made an announcement.

UDS Intrepid

With UDS over it’s time to start the difficult job of deciding what we can actually accomplish for Intrepid.  A partial list of things that might make it in to the Server Edition:

  • J2EE App Server (Full stack or just servlet container remains in question)
  • OpenLDAP based AD proxy
  • Enterprise Management integration
  • Easy LDAP server setup
  • Improved community testing

Some pictures:

Mathiaz enjoying himself at XT3

Howard Chu playing with the Ubuntu All-Stars