Posted by: Jack Vaughan
cloud computing, Data integration
Semiconductor memory advances have powered a new era of portable music and mobile devices. In fact, the general ascent of computer technology has been very much based on cheaper, faster and larger semiconductor memory. The People in the White Smocks are working to ensure that the march continues. Service-oriented architects take note …
Witness this week’s news that IBM and Cal Tech have combined lithographic patterning with self assembly to embed ”DNA origami structures” on surfaces that are compatible with today’s semiconductor manufacturing techniques. This may be an important step forward for chips, as they necessarily get smaller and smaller. [It may also be the biggest thing that happened to origami since the famous Crane-in-a-Turnstile design was formulated in the 1920s! But I digress.]
Ever-cheaper disk memory had a great role in pushing forward SQL relational databases. These DBs still have a long way to go, no doubt. But SQL’s hegemony is questioned these days as vendors, pundits and Joe the Plumber wax poetic on cloud computing and related phenomena. At least one individual sees SQL going the way of Cobol – forgotten but not gone.
Disk memory has made its leaps too. But for clustered Java app servers managing ever-more complex web user sessions, in-memory RAM data caching is gaining ground. ‘Stay away from the disk drive and stay away from the database’ is the mantra of the architects forging low-latency apps. No less than Software AG’s Miko Matsumura has recently pointed to the caching trend:
“…the age of the relational SQL API is coming to a close. Now like any good legacy, SQL will be immortal just like COBOL. But the emerging dominant API will be much more about the network and developer than about the underlying technology. What API better than Java?”
Java as cloud computing language made strides this year as Java support came to both Google’s and Microsoft’s cloud architectures. Both companies have indicated that SQL need not be part of the architecture for data on the cloud – although Microsoft backtracked this summer to add SQL support and placate an army of SQL developers who had quite a few SQL applications working very well before this cloud buzz came along, thank you.
The move to cloud has been seen by some as a great opportunity to say good bye to established data architecture. In some small circles, MapReduce has been much the rage as a new data scheme. But – disk memory, RAM, bubble memory or whatever – there is little doubt that that technology ship is not a vessel that can turn on a dime. Overlaying MapReduce with SQL tools is a way forward, say some.
All this has ramifications for SOA, because – when the smoke clears – SOA is likely to be the lingua franca of the cloud, just as XML was the lingua franca of Web services.
No matter, displacing the old with the new will be a long journey – there is every chance that the conventional architectures will find just enough newness to stay in control – for just a bit longer. What do you think?