Tuesday, January 19, 2010

XTP Processing using a Distributed SEDA Grid built with Coherence

I just finished a talk at the NYC Coherence SIG January 14, 2010.


The topic highlights the concepts Grid Dynamics used to build a high throughput scalable XTP engine for processing telecommunications billing events. The framework employs a distributed SEDA architecture built on top of a Coherence In-Memory-Data-Grid back-end.

6 comments:

pveentjer said...

Hi Terence,

congratulations with your new job. Sorry to see that you and Alex have left tc.

pveentjer said...

Have you ever played with STM?

With STM's a user needs to care less about locking and the programming model is a lot easier than event based (doesn't matter if it is seda or normal.. it is still more complex than 'sequential' code). Of course STM's have their share of problems:
- contention management
- the shared clock a lot of stm's use and this is going to limit scalability and make it hard to distribute the stm.
- not durable.

Currently the company I work for is using gigaspace and although it is a very cool technology, the performance of the teams isn't as high as with a 'classic' approach. This is partly caused by the event driven approach used.

ps:
I'm working on a STM implementation called Multiverse:
http://multiverse.googlecode.com
And I hope to solve some of the issues mentioned above in the near future.

Taylor said...

Peter,

I've been following your work on multiverse and on concurrency-interest list - look forward to your continued progress.

I definitely like the STM idea, although theoretically anyway there are some problems with space under high load.

An area of overlap between an STM approach and the SEDA approach is that the SEDA approach requires a long-running transaction that may span one or more stages.

We solved this problem by making a copy of the data (since it mostly involved shallow objects holding counters and such) and then committed at the end failing and restarting the transaction in the case of a conflict - thus the notes in the preso about an optimistic lock approach.

I think a more general full-featured approach would need to provide options for pessimistic locking, optimistic locking, last writer wins with conflict resolution, etc. provided as configuration.

Nike shox clearance said...

This is a great post,which has a lot of readers.
Such kind of post giving us various knowaledge,it worthy to read.It is so wonderful.

Our company provides Nike Shox Clearance rapid and safe, free shipping network marketing.

Nike shox shoes said...

If you have an expirience with making graphic to web on-line shop please let me know.

Nike shox shoes can surely not let you down,romance yourself with nike shox

aa said...

As I write this post—longhandOffice 2010in a spiral notebook—I’m 20,000 feet above eastern Washington, having Microsoft Office 2010just crossed above the Cascades on my return flight Microsoft wordto Chicago. I visited Seattle for the weekend to Office 2007and I have known each other for 20 years now. They Microsoft Officehad a lovely ceremony, and the trip in general was fantastic.Microsoft Office 2007In the 13 years since I left Seattle, I’ve Office 2007 keyvisited six or seven times, and I always return to wherever has Office 2007 downloadOffice 2007 Professionalbecome home with mixed feelings about the place. It Outlook 2010both alarms and pleases me to see howMicrosoft outlookthat once-familiar areas seem almost foreign. ForMicrosoft outlook 2010neighborhoods have changed, to the point Windows 7 as have cookie-cutter, here-today-and-gone-tomorrow nightclubs that cater to the shiny shirt crowd.