Atomikos for XTP: Trans­ac­tions for Noth­ing and Failover for Free

A lot of the hype in "Ex­treme Trans­ac­tion Pro­cess­ing (XTP)" is fail-over. When Or­a­cle bought Co­her­ence (a Tan­gosol prod­uct), they es­sen­tial­ly got an XTP so­lu­tion for data­base ac­cess.

As Cameron Pur­dy notes here, this now al­lows Or­a­cle to pro­vide a de­gree of XTP failover.

Now guess what: with Atomikos Trans­ac­tion­sEssen­tials you get:

• Trans­ac­tion­al ro­bust­ness for noth­ing, and
• failover for free

How? Just do the fol­low­ing:

1. queue re­quests in JMS
2. process them by a clus­ter of com­pet­ing con­sumer process­es
3. use Atomikos Trans­ac­tion­sEssen­tials to en­sure that each mes­sage is processed ex­act­ly once, with­out du­pli­cates or mes­sage loss

By the se­man­tics of queues, this ar­chi­tec­ture will give you failover. By the se­man­tics of trans­ac­tions, this will give you ex­act­ly once. Since the re­quests can be queued by any source, this is mul­ti­chan­nel. Every­thing is com­mod­i­ty in­fra­struc­ture. This is very easy to scale: just add an­oth­er process.

In sum­ma­ry, this is XTP of the high­est de­gree:-)