One of the themes(/TLAs) mentioned at the Gartner Event Processing Symposium a few weeks back was eXtreme Transaction Processing (XTP) – processing very large amounts of information in a single system / process, instead of a variety of (sometimes mis- or under-) managed stovepipe systems.
Interestingly, this approach has a lot of overlap with CEP:
- read all relevant events into the system (customer transactions, RFID “events”)
- manage all process updates for all main entities “in parallel”
- manage inter- and intra-transaction relationships (such as views by customer, product, department, geography, time, etc)
- provide enterprise-level scaleability and performance, probably in a distributed (multi-geo) environment.
According to our friends at Gartner, “The platform must deal with heterogeneous environments, programmatic and multichannel user access, based on SOA and EDA concepts, integration and process automation, and more.â€
Well, isn’t this what (some) CEP systems can do? I see why. What do you think?
Ophers response blog is at http://epthinking.blogspot.com/2007/11/xtp-distributed-event-processing-and.html
However, CEP tools like TIBCO’s already provide distributable scaleable performance needed for XTP.
For transaction management: well thats probably yet another blog topic: if a customer sends a cancel event, that may have some different consequences for the business transaction based on other events. In other words, also a CEP problem…
Hello Alan,
I’ll take the challange and provide some blog entry about XTP in a few days, need to investigate the concept more before doing that – but in one sentence – it seems that XTP means complete middleware platform that is based on CEP but also other things like grid technology, transaction management etc…
So it seems that XTP needs CEP, but current implementations of CEP don’t qualify as XTP.
cheers,
Opher