Monthly Archives: February 2009
The hardware-vendor cloud
Are hardware vendors the natural winners in cloud infrastructure services?
I don’t think so. Having the lowest cost of server hardware simply isn’t enough. The failure of both Dell and Intel to build successful hosting businesses ought to make that clear. (Dell sold to Sprint, and Intel sold to Savvis, at a significant loss.) Hosting is not an exact parallel, but it’s close enough to look to it for hard-won lessons. Like hosting, cloud infrastructure services are not and will not be just about the lowest costs possible. Cloud will still also be about service, support, and features.
Cloud tidbits in the press
A few tidbits of cloud computing in the press…
Sun announces plans for a cloud computing service. No details until March, other than a ZDnet interview comment about delivering SaaS and SaaS infrastructure.
Rackspace puts its own spin on cloud. Rackspace now has a “cloud 101” page, most interesting for the results of a recent survey that it commissioned on business awareness of “cloud hosting”. No clear definition is provided for that term, though.
Two IBM scientists write a cloud computing article in Dr. Dobb’s. This article made me blink a great deal, starting from the statement, “Currently, you can create cloud applications through two major implementations: Amazon Web Services (AWS) and Google Application Engine (GAE).” The reduction of the richness of the cloud infrastructure services space to just these two names is a little mind-boggling, and the rest of the article that follows is nearly a marketing glossy — accurate in its superficial overview, but offering nearly no information of actual usefulness to an engineer. I hope for better out of technical journals.
Seven years to SEAP, not to cloud in general
Gartner recently put out a press release titled “Gartner Says Cloud Application Infrastructure Technologies Need Seven Years to Mature“, based on a report from my colleague Mark Driver. That’s gotten a bunch of pickup in the press and in the blogosphere. I’ve read a lot of people commenting about how the timeline given seems surprisingly conservative, and I suspect it’s part of what has annoyed Reuven Cohen into posting, “Cloud computing is for everyone — except stupid people.”
The confusion, I think, is over what the timeline actually covers. Mark is talking specifically about service-enabled application platforms (SEAPs), not cloud computing in general. Basically, a SEAP is a foundation platform for software as a service. Examples of current-generation SEAP platforms are Google App Engine, Microsoft Azure, the Facebook application platform, Coghead, and Bungee Labs. (Gartner clients who want to drill into SEAP, see The Impact of SaaS on Application Servers and Platforms.) When you’re talking about SEAP adoption, you’re talking about something pretty complex, on a very different timeframe than the evolution of the broader cloud computing style.
Cloud computing in general already has substantial business uptake, with potential radical acceleration due to the economic downturn. I say “potential” because it’s very clear to me that existing public cloud services, at their current state of maturity, frequently don’t meet the requirements that enterprises are looking for right now. I have far more clients suddenly willing to consider taking even big risks to leap into the cloud, than I have clients who actually have projects well-suited to the public cloud and who will realize substantial immediate cost savings from that move.
On the flip side, for those who have public-facing Web infrastructure, cloud services are now a no-brainer. Expect cloud elasticity and fast provisioning to simply become part of hosting and data center outsourcing solutions. Traditional hosting providers who don’t make the transition near-immediately are going to get eaten alive.
Google Federal
I heard a radio ad today for Google Federal. It sounded like every other “please, government IT purchasing person, buy our stuff” ad that you hear on news radio in Washington DC. It was a far cry from the sort of ad that one expects to hear from Google, and to hear a federal-targeted ad from them, period, was sort of fascinating.
The Federal government can still afford stuff (and is probably one of the few bright spots in purchasing this year, period). It’s the states that are screwed, and seriously thinking about alternatives to traditional IT.
Volume pricing for Amazon’s CloudFront
New volume pricing for Amazon’s CloudFront CDN takes effect today, February 1st. For US and Europe “edge” delivery, the price goes as low as $0.05/GB at the 1000+ TB level. For Hong Kong, it’s $0.09/GB at that level. For Japan, $0.095/GB. The pricing isn’t quite comparable to a traditional CDN because of the origin bandwidth fees and the per-request fee, but it’s still a useful benchmark.
For those who are mentally comparing this to the cost of bandwidth, those per-GB costs translate into $16/Mbps for US/Europe, and $29/Mbps for Asia. In a day and age when Cogent is splashing “Home of the $4 Megabit” across its home page, it might look like there’s still quite a bit of delta between bandwidth pricing and CDN pricing, but especially once you get out of the US, bandwidth costs escalate pretty dramatically beyond Cogent’s low-water-mark.
Nonetheless, Amazon’s volume pricing play ought to put to an end anyone’s hope that the elimination of some of the financially weaker CDN players is going to do anything significant to alleviate pricing pressure where it’s most severe — the entirely commoditized portion of the market. In fact, this explicit, transparent pricing is probably going to provide a nice bargaining chip. Even if a major media conglomerate isn’t going to use Amazon to deliver their video, it won’t stop its purchasing people from using these published prices to hammer CDNs during negotiations.
CloudPundit: Massive-Scale Computing 