With the latest version of its Enterprise Application Fabric (EAF), Appistry seems to have found a niche for itself somewhere at the intersection between real-time infrastructure (RTI) and Grid computing, said Carl Claunch, vice president of research and advisory services at Gartner.
But what makes Appistry's approach to fabric solutions unique, and what, exactly, is a fabric?
To answer the latter question, Claunch said at Gartner, “Our review of Grid is that it is to take a single task and accelerate it by running it across multiple resources.
“A fabric, on the other hand, is a more generalized approach to running any kind of workload — not just queuing up one task, but running a mixture of unrelated work. We call that 'real-time infrastructure' when we talk about this approach, and we think it is the most significant, widely applicable approach.”
Kevin Haar, CEO of Appistry, is not so strict in his definition of a fabric, but he does acknowledge that the real-time functionality is definitely a key selling point of the company's EAF solution. “What we really are trying to do is bring a notion of real-time capability to Grid,” he said, “and we like to call it a fabric, but we don't get mad if people call it a grid.”
Haar continued to discuss the difference, noting that for customers, “A big barrier has been … the ability of them to apply Grid to the types of problems they want to solve. One of the big issues there has been the notion of a batch-orientation versus more of a real-time, transactional type of orientation.”
So, if we can agree that fabric solutions generally play more in the RTI realm than do traditional Grid solutions, we are left with the question of what makes Appistry's solution so unique. After all, as Claunch pointed out, if the company were competing solely in the RTI space, it would be “paddling hard against the current” as it fought to compete with some big players and some quality solutions, such as IBM with its On-Demand initiative and HP's Adaptive Infrastructure. As for the Grid space, Appistry would again be competing with more established vendors (e.g., United Devices, Platform and DataSynapse) with decent solutions. However, as Claunch likes to point out, Appistry is best at the intersection of these two trends.
In EAF 3, the company has introduced new workload management and execution policies that make it more conducive to grid-type jobs than were previous versions. Haar said that Appistry has been focusing on closing the gap between computational and transactional applications, and while previous versions of EAF had load-balancing algorithms, etc., to distribute work, they were tuned toward smaller units of work.
With the new execution policies, however, Claunch believes Appistry can fill one of the “maturity gaps” that exist with traditional Grid tools. “The biggest challenge for people who are trying to do Grid is when they really do have tight service-level requirements,” he said.
“Traditional Grid tools are great for harvesting the raw performance, but they can't guarantee very consistent turnaround times, whereas Appistry has that capability.”
Another thing that makes Appistry's solution unique is its use of virtualization technology, which Haar believes makes it very simple to manage versus other similar solutions. “We talk about scale-out virtualization,” he said, “and what we mean by that is just an opportunity to have a very scaled out, grid-like architecture, but it looks like one thing to the developer and one thing to the operations group.”
Claunch agreed with this assessment, adding that Appistry does a good job of letting the application or the person managing the fabric think of the resident machines as a cloud. “Whether it's running on one or 50 machines is sort of hidden, almost entirely, from the application,” he said. “Whereas with other Grid solutions, you tend to develop the application to say it's going to run on 50 machines at one time.”
Speaking of machines, the hardware infrastructure is another area where Appistry believes it has an edge over the competition. Sam Charrington, the company's vice president of product management and marketing, said EAF offers more of a data center-oriented capability, and that Appistry isn't “trying to cycle-scrape off desktops or that sort of thing.” What Appistry is trying to do, however, is allow customers to be as aggressive about commodity as they like, Charrington said. Whether customers want to use blades, 1Us or commodity white boxes, “Our take on it is: you use the fasted thing that's cheap.”
The reason for this open attitude toward hardware is EAF's application-level failover, which Charrington said means users are no longer dependent on hardware reliability. Haar further praised EAF's fault-tolerance, saying that by providing full failover, “not only can we predict when something will get done, but we can make sure that it did, in fact, get done.”
As for what kinds of customers Appistry is seeing for its EAF solution, Haar said one interesting customer uses EAF to manage satellite imagery. The imagery work is very CPU-intensive, but the customer needs to be able to use the data generated from its satellite imaging application in a near real-time setting. Another image-processing customer, Haar said, is using EAF to run a Web application that delivers personalized marketing services and images to its clients. Due to the amount of computing resources needed to generate the images, as well as the real-time necessity of a Web application, this customer is really benefiting from the new execution policies.
Aside from image-processing customers, Haar said Appistry is also being used by customers in the intelligence community, as well as by users in the fields of transportation and retail, and the company is making inroads in the fields pharma, genetics and bioinformatics.