Real-time computing. Real-time news.
June 16, 2008
Mission-Critical Cloud Computing? Check Back in Five Years
1 | 2 | 3 All »
Let's fast-forward through the question of when enterprise IT will fully, heatedly embrace cloud computing -- the answer, according to analysts, is five years -- and get right to the better question: Will they ever trust the cloud with their life-or-death applications? Those big, demanding applications that require what some call extreme transaction processing -- trading, reservations, electronic payments, etc. -- can you run those in the cloud? Do you want to?
GigaSpaces Technologies specializes in helping companies develop distributed, scalable, on-demand systems that can handle big, honking, rapid-fire enterprise and Web applications. And they intend to help companies run those systems in Amazon's Elastic Compute Cloud.
But before we go to the cloud, a bit of background about GigaSpaces' approach to delivering scalable applications. GigaSpaces's flagship product is an application server built from scratch for intense computing. The name, eXtreme Application Platform (XAP), kind of gives that away. The company describes it as middleware for running high-performance, high-reliability applications on grids and other distributed systems. The biggest challenge that XAP is designed to tackle is scalability now.
"It's not the constant growth in the amount of data, transactions and service requests. It's the unpredictable peaks and troughs," says Geva Perry, chief marketing officer at GigaSpaces. "Like when AT&T had to provision all those iPhones suddenly for more people than expected and had systems crashing." That might also be a business modeling problem, but the point is those peaks can be erratic, and the one-time events will get you every time. (Nobody expects the Spanish Inquisition.)
"You can throw money at the problem and just buy lots of servers to have on hand," Perry says. "But a lot of companies do that and realize they've overprovisioned and have all that stuff sitting idle." Underprovisioning can have worse consequences.
"What we have is an application platform that allows you to scale cost-effectively and quickly on demand with no changes to your application," Perry says. "XAP lets you build a high-throughput application so that as demand grows, it can respond, and you don't have to use any new APIs or make any architecture changes. Developers can write in their usual Java or .Net or whatever."
"XAP enables your applications to scale linearly, and that's the only way to scale effectively," Perry says. "Add 100 servers, then handle 100 times more transactions. But the reality is most middleware products don't handle things the way we do and end up with bottlenecks in one system or another. Doubling servers doesn't double throughput." In more painful accounting terms, we're talking about diminishing return on hardware investment.
GigaSpaces says it does a few things differently that avoid dreaded latency. All services reside in the same server, eliminating the usual hops between the messaging system, the database, and so on. "With XAP, a transaction's data, business logic, and messaging are all completing in the same place," Perry says. In the GigaSpaces universe, applications travel as self-sufficient "processing units." When it's time to scale up to meet demand, you add more processing units. "It's simple scaling. One click," Perry says. XAP also executes every transaction in local memory, avoiding a trip to the database server (transactions are archived there later).
GigaSpaces says XAP's scalability and performance features meet the needs of large-scale applications, including SaaS, financial services, e-commerce, online reservations, telecomm provisioning and gaming, and that it has customers in all those areas. Financial services company Susquehanna International Group, for example, built its distributed trading platform, which relies on multiple low-latency applications, on top of XAP.
Persistent Clouds
Page: 1 of 31 | 2 | 3 All »
Article Tools
Share Options
(Digg, Technorati, more)
