|
|
AS OF 5/17/2008 12:27PM EST
|
Rogue Wave Puts Teeth of Hydra Upon SOA
Parallelization boosts speed by breaking applications into smaller bites
By Andy Patrizio
March 15, 2006 —
Rogue Wave is promising to make service-oriented architectures run an order of magnitude faster thanks to Hydra, a framework that it says makes the sequential processes of SOAs run in parallel.
Hydra uses Rogue Wave’s Pipelines technology to break the steps in SOA-based applications into scalable processes that can be operated in parallel, without waiting for slower steps in the process to complete, a frequent problem in the loosely coupled SOA world.
“SOA is primarily based on loose coupling, using XML as a transport,” said Cory Issacson, president of Rogue Wave. “While that’s very flexible and nice, it’s inherently slow. So how do you make it fast?”
Parallel processing is hardly new, but as a concept it’s been used largely in what Issacson called “embarrassingly parallel problems,” such as fluid dynamics and financial analysis applications. For example, a parallel process would divide a process that normally would take 100 hours to complete into 100 pieces to run on 100 computers. The results are assembled at the end. That may work nicely for SETI@Home (the grid computing project that helps search for evidence of radio transmissions from extraterrestrial intelligence), Issacson said, but not for an ATM transaction system.
Business applications have requirements like first-in, first-out ordering that make parallelization difficult. Rogue Wave’s Pipelines allow developers to control the order of their processes while dividing them up to avoid bottlenecks, said Issacson. For example, a bank ATM processor would handle millions of transactions per day, if it’s a large bank with many ATMs. Hydra would allow the bank to split its processes, such as authentication, transactions, processing and settlement, into separate processes that don’t necessarily run in sequence.
However, don’t expect Hydra to be a magic bullet for parallelization; infrastructure will play a big role in successful use of Hydra. “Execution takes a huge amount of understanding. What you have might not be componentized the right way to take advantage of parallelism,” said Issacson. A large, monolithic program to run a complex process, for example, can’t magically run in parallel on 100 CPUs. The application may have to be partitioned into tasks to be properly parallelized, he said.
To help with this process, Rogue Wave and its partners are offering a Pipelines Analysis service, where they analyze a customer’s applications and determine how well they will fit into the pipeline methodology.
Rogue Wave Hydra 2.0 is available now for Linux and Windows; pricing was not disclosed.
|