Server maker Sun Microsystems will today launch its third assault into utility-style computing. And if you find yourself having a flashback of sorts, it isn't you and all that brown acid you took at Berkeley. It's just the way it is in the modern computer business.
Ask anyone in IT, except for maybe a few million CIOs and business owners who have their skeptical caps on, and they will tell you: Cloud computing is an idea whose time has come - again. And if at first you don't succeed, that doesn't mean it was a bad idea. It is just that someone else has done the idea better and you have to catch up so you can get some of the cash companies are going to be throwing around to build these computing and storage infrastructures called clouds. I think the old-fashioned term "utility" is better suited to what is really happening, but that is so 20th century, as is "grid," the term Sun preferred way back when the Sun Grid was launched in February 2005. Humans are so fickle. Don't blame Fate's finger.
Today, at the CommunityOne East developer conference Sun is hosting in New York, the company will announce the Sun Cloud, which Sun is billing as the "first public cloud service offering for developers, student, and startups," a cloud that will feature compute and storage services. I guess by first that means if you don't include Sun Grid or its kicker, Network.com, which was aimed at developers and startups as far as I can remember as well as corporations looking to offload some Solaris work to Linux or Solaris machines (that was the Sun Grid) and then only Solaris (that was Network.com).
Whatever. The point now, according to Juan Carlos Soto, vice president of cloud computing marketing at Sun, who has run its marketing program to chase startups as well as being chief technology officer for Sun's software business, is that Sun is going to operate a public cloud and will initially target developers and startups as the users for the platform. These are the kinds of customers who don't have big capital budgets and would no doubt prefer to rent compute and storage capacity than own gear and pay upfront for it.
This time around, according to Soto, Sun is building a public cloud that will support Linux, Windows, and Solaris - not just Solaris, which kind of limited its appeal - on a mix of Sparc and x64 iron. The cloud will be built using blade servers, which means Niagara processors on Sparc blades and both Xeon and Opteron processors on x64 blades. Because the initial targets are developers and startups, Solaris 10 is not going to be available out of the chute on the clouds, but the OpenSolaris development distro will be on both Sparc and x64 chips.
Several flavors of Linux and Windows will also be available on the x64 iron. To virtualize this server iron, Sun will be using its own xVM hypervisor tools, which of course includes Sun's implementation of the open source Xen hypervisor on x64 machines and logical domains (LDoms) on the Niagara iron. Both x64 and Sparc versions of the Solaris operating system can also support Solaris containers, which is Sun's riff on virtual private servers.
Down the Amber Road
These blades will be backed up by Sun's "Amber Road" open storage arrays, which sport Solaris 10 and ZFS, among other things. These arrays do not have flash-based solid state disks in them yet, but very well could soon, since Sun is trying to get developers interested in these SSDs so they can talk their companies into buying them. Soto says that the storage part of the cloud will have an S3-like service that is "conceptually more like block-level storage" and that Sun will eventually offer file-level access to the storage service using the WebDAV protocol.
The secret sauce to the Sun Cloud, according to Soto, is not all of the iron and virtualization, but the interface that Sun has created to allow developers and startups to create a "virtual data center" using drag-and-drop tools that allow them to create a set of software stacks - for load balancers, Web servers, proxy servers, database servers, and such - and just pull them from the library and plunk them out onto the real Sun cloud. Sun is interested in supporting the open source Hadoop distributed computing environment on the cloud, of course, and will be layering other services on top of the compute and storage services. A MySQL database service is an obvious one.
As we reported last week, Sun's chief technology officer, Greg Papadopoulos, had already outed the Sun Cloud, saying in a keynote address at the AFCOM Data Center World conference that Sun's next-generation cloud was being housed at the SuperNAP data center compound run by Switch Communications outside Las Vegas.
Soto says that Sun has had the Sun Cloud up and running for the past few months and that a number of alpha customers have been putting it through the paces. Presumably, the Sun Cloud will be as inexpensive and easy to use as the touchstone in this cloud racket, Amazon's Elastic Compute Cloud (EC2) and Simple Storage Service (S3). While Sun is launching the Sun Cloud today, and its Cloud Compute Service and Cloud Storage Service (with no funky abbreviations like C2S or CS2), it is not announcing pricing. Those will be announced some time this summer, when the Sun Cloud becomes generally available, as will detailed service level agreements.
So, can Sun eventually make money at this cloud thing? "Not only do we believe that this will be a profitable business," says Soto, "but that the time is now right to enter the cloud computing market."
If that is the case, Sun had better set its targets a lot higher than developers, students, and startups. They don't cut the big checks or cheques, depending on what side of the Pond you are on. To be sure, you need to include these types of users, but that is not where the money and profits are going to be in this down economy. Companies want to cut IT budgets, and anyone who can demonstrate that their cloud is as good or better than whatever the IT department can build - and do so for less money - is going to get some business. ®