SQLAzure LittleBigPaaS with Michael Rys
Best with last evenings Microsoft SQL Azure event at the Seattle Hadoop meetup was Michael Rys live demo coding how SQL Azure’s splits shards for elastic scaling, and the state of folding shards back into main without losing any consistency at all. Essentially by taking SQL and NoSQL learning to date and architecting the Azure Platform to relax the cap theorem.
The principal business impact I am concerned with is the ability of big data business apps to elastically scale up and back down again. Ideally without DevOps or IT interventions.
Currently the cloud biz apps platform market leader, both on revenues and thought leadership is Salesforce.com. Setting a context for comparing SQL Azure and Force.com platforms. Compare say a ticketing application to handle scalability needs when Lady Gaga tickets go on sale. (remember the Google IO 2011 tix slashdotting effects).
Last nights audience drawing heavily from the Hadoop and NoSQL community. Mike pointed out what’s obvious to this community, scaling has mostly been achieved by relying on applications to scale. Not on platforms. Yet, many client sever apps scale poorly, often topping out at 300 concurrent users or less. Big data biz apps need to scale to 30,000 or more.
By learning internally and from observing the NoSQL vendors, SQL Azure has been engineered as a platform for scaling, rather than relying on applications to do the scaling. Thus putting more power in the hands of smaller development teams without rock star architects experienced at scaling 1 to 30,000 (and to 300,000…)
The Force.com platform, engineered a solid 10+ years ago on database-on-a-database pattern locks scalability into the rate limiting allowed in the multitenancy layer. For business developers suffocating in IT bureaucracy, hardly a prison cell, still a prison (still a real money maker for Salesforce.com).
The next version of Java is slated to include capabilities for multitenant applications to sense database loading presumably to allow for provisioning additional capacity from underlying Oracle SQL database. This may allow Force.com architect Parker Harris to re-engineer and remove rate limiting. But what incentives does Oracle have to make cloud applications big data scalable, when the primary Oracle market is on-premise enterprise applications?
Where as Microsoft Windows platform likely having achieved peak revenues, has huge incentives to provision a Cloud PaaS platform for developers to recreate new little and big business applications with scaling provisioned by the platform not in the application itself. Ushering in the PaaS Developers.
A double check on this strategy Saleforce.com has shifted up the application stack to the Social Enterprise applications on the Heroku polyglot platform.
Microsoft SQL Azure perhaps doesn’t push Force.com aside, Salesforce already moved.
(Source: meetup.com)
![Murdoch on MySpace: “We screwed up in every way possible
Facebook was quick to embrace Ajax but MySpace did not follow suit, partly because to do so would have reduced the number of page views the site generated and therefore its advertising revenue. “It would take five steps to post a comment or send a message, so five different pages would open,” explains another former executive. “There would be ads on each of those pages, so we were making money. We went to News Corp and said: ‘We want to change this but in the short term our revenues will drop.’ It became a long back and forth. [Newscorp overseers] were pushing back – they wanted to make sure we weren’t going to drop our revenue numbers.”](http://27.media.tumblr.com/tumblr_lxu5x1ImH21qz7jv1o1_500.png)
