Long Transactions and Slack

One of the most contentious problems in Web Services and online transactions is long-duration transactions. Now long duration is not microseconds or even seconds; but rather think in terms of minutes to hours to days. Think in terms of long duration Web Transactions or messaging workflows or various replication strategies. Now not all of these trasactions are being done under ACIDs commit and rollback conditions; but still long transactions present some formidable challenges for correct processing.

The reason this is of great import is the new wave of Mobile and workflow based applications as envisaged in SOA-Service Oriented and especially ESB-Enterprise Service Bus frameworks and arhitectures. This is important because many of the strategies, resources and simple slack (inventory, downtime, extra-capacity, etc) needed to resolve the inevitable conflicts and mismatches engendered by long transactions have been relentlessly weeded out of systems in the first phase of development – when transaction were primarily instantaneous. So system development is going to have to reprioritize yet again. It is going to have to unlearn some of what it has so recently mastered. This is just another factor that will make Progress(read n-tier heterogeneous processing using SOA and ESB), which is our most important product, every bit as socio-economically challenging as it is inherently complex.

(c)JBSurveyer 2005