cloud computing | Ted Leung on the Air

On Tuesday I attended Google’s Chrome update event in San Francisco. There were three topics on the agenda: Chrome, the Chrome Web Store, and ChromeOS. I’m not going to try to go over all the specifics of each topic. It’s a pointless exercise when Engadget, PC Magazine, etc are also at the event and live blogging/tweeting. I’m just going to give some perspectives that I haven’t seen in the reporting thus far.

Chrome

If you are using a Chrome beta or dev channel build, none of the features announced would be new to you. The only exception is the Crankshaft technology that was added to V8. The claim is that Crankshaft can boost V8 performance up to 50%, using techniques which sound reminiscent of the HotSpot compiler for Java. Unsurprising that the V8 team includes veterans of the HotSpot team. Improving Javascript performance is good, and in this case it’s even better because V8 is the engine inside Node.js, so in theory Node should get some improvements on long running Javascript programs on the server. I’m pretty sure that there is some performance headroom left in Crankshaft, so I’d expect to see more improvements in the months ahead.

The Chrome team has the velocity lead in the browser wars. It seems like everytime I turn around Chrome is getting better along a number of dimensions. I also have to say, that I love the Chrome videos and comic books.

Chrome Web Store

So Chrome has an app store, but the apps are websites. If you accept Google’s stats, there are 120M Chrome users worldwide, many of them outside the US, and all of them are potential customers of the Chrome Web Store, giving it a reach comparable to or beyond existing mobile app stores. The thing that we’ve learned about app stores is that they fill up with junk fast. So while the purpose of the Web Store is to solve the app discover problem (which I agree is a real problem for normal people), we know that down that path lie dragons.

The other question that I have is will people pay to use apps which are just plain web apps? Developers, especially content developers, are looking for ways to make money from their work, and the Chrome Web Store gives them a channel. The question is, will people pay?

ChromeOS

The idea behind ChromeOS is simple. Browser as operating system. Applications are web applications. Technically there are some interesting ideas.

The boot loader is in ROM and uses crypto to ensure that only verified images can be booted (the CR-48 has a jailbreak switch to get around this, but real hardware probably won’t). It’s the right thing to do, and Google can do it because they are launching a new platform. Is it a differentiator, maybe if you are a CIO, or a geek, but to the average person this won’t mean much.

Synchronization is built in. You can unbox a ChromeOS device, enter your Google login credentials and have everything synced up with your Google stuff. Of course, if you haven’t drunk the Google ecosystem Cool-Aid, then this won’t help you very much. It’s still interesting because it shows what a totally internet dependent device might be like. Whatever one might say, Android isn’t that, iOS isn’t that, and Windows, OS X, and Linux aren’t that. When I worked at Sun, I had access to Sun-Ray’s, but the Sun Ray experience was nowhere as good as what I saw yesterday.

There’s also some pragmatism there. Google is working with Citrix on an HTML5 version of Citrix’s receiver, which would allow access to Enterprise Applications. There are already HTML VNC’s and so forth. The Google presenter said that they have had an unexpectedly large amount of interest from CIO’s. Actually, that’s what led to the Citrix partnership.

Google is piloting ChromeOS on an actual device, dubbed CR-48 (Chromium isotope 48). CR-48 is not for sale, and it’s not final production hardware. It’s a beta testing platform for ChromeOS. Apparently Inventec (ah, brings back my Newton days) has made 60,000 devices. Some of those are in use by Googlers, and Google is going to make them available to qualified early adopters via a pilot program. The most interesting part of the specs are 8 hours of battery life, 8 days of standby time, and a built in Verizon 3G modem with a basic amount of data and a buy what you need for overages.

Hindsight

At the end of the presentation, Google CEO Eric Schmidt came out to make some remarks. That alone is interesting, because getting Schmidt there signals that this is a serious effort. I was more interested in the substance of his remarks. Schmidt acknowledged that in many ways, ChromeOS is not a new idea, harking back (at least) to the days of the Sun/Oracle Network Computer in the late 90’s. In computing timing matters a huge amount. The Network Computer idea has been around for a while, Schmidt claimed, but it’s only in this day, that we have all of the technology pieces needed to bring it to fruition, the last of the pieces being a version of the web platform that is powerful enough to be decent application platform. It’s going to be interesting to see whether all the pieces truly have arrived, or whether we need a few more technology cycles.

The question

Last week there were some articles, blog posts and tweets about the relationship between Platform as a Service (PaaS) offerings and open source. The initial framing of the conversation was around PaaS and the LAMP (Linux/Apache/MySQL/{PHP/Perl/Python/Ruby}) stack. An article on InfoQ gives the jumping off points to posts by Geva Perry and James Uruqhardt. There’s a lot of discussion which I’m not going to recapitulate, but Uruqhardt’s post ends with the question

I’d love to hear your thoughts on the subject. Has cloud computing reduced the relevance of the LAMP stack, and is this indicative of what cloud computing will do to open-source platform projects in general?

Many PaaS offerings are based on open source software. Heroku is based on Ruby and is now doing a beta of Node.js. Google’s App Engine was originally based on Python, and later on Java (the open sourceness of Java can be debated). Joyent’s Smart Platform is based on Javascript and is open source. Of the major PaaS offerings, only Force.net and Azure are based on proprietary software. I don’t have hard statistics on market share or number of applications, but from where I sit, open source software still looks pretty relevant.

Also I think it’s instructive to look at how cloud computing providers are investing in open source software. Rackspace is a big sponsor of the Drizzle project, and of Cassandra, both directly and indirectly through its investment in Riptano. EngineYard hired key JRuby committers away from Sun. Joyent has hired the lead developer of node.js, and VMWare bought SpringSource and incorporated it into VMForce. That doesn’t sound to me like open source software is less relevant.

Cloud computing is destined to become a commodity

The end game for cloud computing is to attain commodity status. I expect to see markets in the spirit of CloudExchange, but instead of trading in EC2 spot instances, you will trade in the ability to run an application with specific resource requirements. In order for this to happen, there needs to be interoperability. In the limit, that is going to make it hard for PaaS vendors to build substantial platform lockin, because businesses will want the ability to bid out their application execution needs. Besides, as Tim O’Reilly has been pointing out for years, there’s a much more substantial lock in to be had by holding a business’s data as opposed to a platform lock. This is all business model stuff, and the vendors need to work this out prior to large scale adoption of PaaS.

Next Generation Infrastructure Software

The more interesting question for developers has to do with infrastructure software. In my mind LAMP is really a proxy for “infrastructure software” If you’ve been paying any attention at all to the development of web application software, you know that there is a lot happening with various kinds of infrastructure software. Kiril Shenynkman, one of the commenters on Geva Perry’s post wrote:

Yes, yes, yes. PHP is huge. Yes, yes, yes. MySQL has millions of users. But, the “MP” part of LAMP came into being when we were hosting, not cloud computing. There are alternative application service platforms to PHP and alternatives to MySQL (and SQL in general) that are exciting, vibrant, and seem to have the new developer community’s ear. Whether it’s Ruby, Groovy, Scala, or Python as a development language or Mongo, Couch, Cassandra as a persistence layer, there are alternatives. MySQL’s ownership by Oracle is a minus, not a plus. I feel times are changing and companies looking to put their applications in the cloud have MANY attractive alternatives, both as stacks or as turnkey services s.a. Azure and App Engine.

How many of the technologies that Shenykman lists are open source? All of them.

Look at Twitter and Facebook, companies whose application architecture is very different from traditional web applications. They’ve developed a variety of new pieces of infrastructure. Interestingly enough, many of these technology pieces are now open source (Twitter, Facebook). Open source is being used in two ways in these situations. It is being used as a distribution mechanism, to propagate these new infrastructure pieces throughout the industry. But more importantly (and for those observing more closely, quite imperfectly), open source is being used as a development methodology. The use of open source as a development methodology (also known as commons-based peer production) is definitely contributing to these innovative technologies. Open source projects are driving innovation (this also happened in the Java space. Witness the disasters of EJB 1.0 and 2.0 which lead to the development of EJB 3.0 using open source technologies like Hibernate, and which provided the impetus for the development of Spring). Infrastructure software is a commons, and should be developed as a commons. The cloud platform vendors can (and are) harvesting these innovations into their platforms, and then finding other axes on which to compete. I want this to continue. As I mentioned in my DjangoCon keynote last year, I also want open source projects to spend more time thinking about how to be relevant in a cloud world.

My question on PaaS is this: Who will build a PaaS that consolidates innovations from the open source community, and will remain flexible enough to continue to integrate those innovations as they continue to happen?