This is an interesting post that I’ve read several times. My thought and understanding can be followed by highlight in bold.
At the LinuxWorld and Next Generation Data Center Expo in San Francisco, Merrill Lynch’s chief technology architect, Jeffrey Birnbaum, discussed stateless computing and the evolution of the cloud. He envisions a future in which software processes are abstracted so far from the underlying hardware that companies will discuss processing capacity in terms of raw computational units rather than discrete servers.
The real underlying value of cloud computing, he says, is that it transparently makes software and data available everywhere. He contends that this “stateless” model facilitates much greater scalability than conventional computing and can be used in conjunction with virtualization to achieve maximum data center utilization. “Stateless will emerge as a core, basic tenet of what’s in the cloud,” he remarked.
During his presentation, he described the architecture of a globally distributed stateless software implementation and explained some of the steps that Merrill Lynch has taken to move towards this model for its own internal IT infrastructure.
Birnbaum believes that one of the key foundational elements of a stateless computing environment is a networked storage system that enables ubiquitous availability of software. The file paths of the individual applications should be based on clearly defined nomenclature, much like the domain of a web site. All application dependencies should be accessible through the network filesystem, and version numbers should be expressed with the path nomenclature.
There are numerous advantages to this kind of network storage scheme. In addition to making the information and software universally accessible, it also obviates many of the technical challenges that are typically associated with software deployment. He also dubiously claims that this approach eliminates the need for preintegrated software stacks, a claim that I view with great skepticism (I’d argue that a big part of preintegration is harmonizing disparate components, and that entails a lot more than just deployment).
The obvious challenge posed by rolling out worldwide network storage infrastructure is scalability. If everyone in a global organization is depending on a network storage solution, then it needs to be fast and consistently reliable. The solution that Birnbaum proposes is regional mirroring and caching. The storage system would be universally synchronized between mirrors that have all the data. Caching can also be used at individual facilities to further improve performance. To achieve this kind of global scalability, he says, the best approach is similar to that of Akamai.
The system Birnbaum describes isn’t just theoretical. Merrill Lynch has been deploying its own implementation, which they call the Enterprise File System (EFS). Applications are streamed across the network and operate with no persistent state on local systems. This has worked well with Linux-based applications over NFS, but he says that they have encountered some challenges with Windows, which wasn’t designed to be used in that manner. For Windows applications, he suggests using virtualization—the user employs RDP to access applications running in virtualized environments in the data center.
These concepts don’t cover a whole lot of new ground yet. Much of this was already possible with conventional thin-client systems. The point at which it becomes immensely valuable, according to Birnbaum, is when all of these technologies are used together with virtualization to abstract the processes away from the hardware. Once this is done, individual operations can seamlessly float around data centers and balance out in a manner that offers a more optimal level of resource utilization.
He claims that 61 percent of a company’s enterprise server capacity goes completely unused and proposes an automated load balancing solution—a placement engine—which will manage the resources and the active processes, moving them between servers so that the existing capacity can be used more efficiently. His goal is to reach approximately 80 percent utilization (he cautions against going higher than that because he thinks it’s beneficial to always have spare capacity for unexpected circumstances).
One of the big advantages of this approach is that it makes it possible for companies to use cheaper hardware. Each individual server doesn’t need to be extremely reliable because processing can always be moved elsewhere in the event of hardware failure. The placement engine, however, is still the missing piece of the puzzle. Virtualization management technologies just aren’t smart enough to do that kind of manipulation in a fully automated way yet. He also says that the industry needs “bigger pipes and lower-latency pipes” in order to handle the constant flow of data (we wouldn’t want the tubes to clog).
The ideas he presented reflect several emerging IT trends. Virtualization is becoming increasingly prominent in data centers because of its advantages for utilization and scalability. The underlying message of his presentation is that the central principles of cloud computing can be adapted for use with conventional software and leveraged to increase the overall efficiency of computation. These ideas could help shape the way that next generation data centers are architected.
