Article: Java Grid, why do we need it?

Posted by Floyd Marinescu on Oct 12, 2007 09:32 AM

Community

Java

Topics

Performance & Scalability,

Grid Computing

Tags

GigaSpaces,

Terracotta,

Tangosol,

Jini

Investment banks have often set trends or offer the rest of us a glimpse into the future of our IT infrastructure. Today banks are investing heavily into grid technologies as they continue in their battle for profits. In this article, John Davies shares his experiences on technologies used in finance and why Grid technologies are becoming a new mainstream for high volume, low latency systems.

Read Java Grid, why do we need it!

John concludes:

The clever vendors are now programming to a grid culture, simple thread-safe POJOs with self-contained functionality and defined behaviour, you and I know these as objects. As we strive to get every last cycle out of the CPU we're looking to make the most out of the CPU's direct memory. Distribution is inevitable but like the way a CPU's memory cache works we try to optimise the network usage to a minimum, advances in CPU technology and the ever-diminishing price of memory is playing right into our laps. Grid, however you define it, is here to stay.\

What have been your experiences with Grid archittectures?

1. Back to top

   Paremus Infinflow

   Oct 11, 2007 5:43 AM by richard nicholson

   John, You fail to mention Infiniflow. Industry's only OSGi / SCA based distributed self-healing service fabric. In addition to providing traditional Grid type behavior for Spring and Java POJO based applications. Infiniflow replaces the very concept of Application Service and Messaging Middleware - with a dynamic OSGi based service based runtime. Come and see our presentation at NYJavaSig next week. Cheers Richard

   Reply

2. Back to top

   A couple other things

   Oct 11, 2007 9:23 AM by Cameron Purdy

   See also: http://gridgain.com/ Peace, Cameron Purdy Oracle Coherence: Data Grid for Java and .NET

   Reply

3. Back to top

   Re: A couple other things

   Oct 11, 2007 2:28 PM by Nati Shalom

   Great writeup John (As always)

   Now as soon as this is posted I know each of the players will comment saying that theirs is the best over all product and that the others are simply crap. Each will be able to dig up examples of where they've won over the competition and each will be able to demonstrate holes in the other's architecture but at the end of the day they all work well if you chose the right tool for the problem

   I''ll save you and the rest of the community that pleasure this time:) I would however point you to a resent discussion on this topic here and here which provides further clarifications on GigaSpaces position as it relate to Grid, Scalability, Caching etc. Nati S. GigaSpaces Write Once Scale Anywhere

   Reply

4. Back to top

   Few comments

   Oct 12, 2007 3:19 PM by Nikita Ivanov

   Interesting article. Just proves how views on grid computing can be different (not worse – just different). I, for example, would never put Gemstone into the mix of grid vendors. In the same time, not mentioning entrenched vendors like DataSynapse, Platform Computing and United Device/Univa is rather strange. Mentioning of Globus Toolkit cannot be found either. Another point is that for me GigaSpaces is much more of a data grid than a compute grid (master/worker has nothing to do with either one – it can be implemented in any product what-so-ever). Yet another point is that Terracotta’s strength is not in open source license (which requires attributions in its case, btw) but in unique technological approach of clustering low-level Java semantics (like synchronization, etc.). More over, GigaSpaces too has open source option (OpenSpaces) albeit limited. Having certain actual experience in this area I can say that the key characteristic of any mature *compute* grid product is rich support for Map/Reduce logic (which is an essence of parallel computing in general dating back to MPI time). Recent fascination and confusion about Master/Worker pattern in relation to grid computing is rather startling: since when an old rusty RPC became a staple of grid computing :-) Best, Nikita Ivanov. GridGain - Grid Computing Made Simple

   Reply

5. Back to top

   Re: Few comments

   Oct 12, 2007 4:20 PM by Cameron Purdy

   Recent fascination and confusion about Master/Worker pattern in relation to grid computing is rather startling: since when an old rusty RPC became a staple of grid computing :-)

   Since JavaSpaces vendors had to find a way to be "grid" vendors! (Sorry .. I couldn't resist .. ;-) Peace, Cameron Purdy Oracle Coherence: Data Grid for Java and .NET

   Reply

6. Back to top

   Re: Few comments

   Oct 13, 2007 1:36 AM by Julian Browne

   Why is it than whenever anything related to grid technology gets mentioned, either here, TSS, Artima. There has to be this vendor FUD sown about? I don't remember any other solution area suffering this way so early in the adoption cycle. John's article was one of the best, and balanced, I've seen on the state of things in a long time. From a customer perspective, it's hard to put a value on how useful this is. We're still fighting a rear-guard action against point-to-point integration, propping up badly done EAI from the nineties, having to sift through SOA-bollocks from 'consultancies' (while the board continually asks if it's done yet), rationalising ESB use, and now worrying about EDA and XTP. As an architect it's hard enough convincing the business that any approach is worth the extra effort in these circumstances, and if the convincing takes too long, the developers have usually gone off and dropped another tactical solution into operations for us to worry about next year. What adds wonderfully to this is vendors redefining concepts and pointing out perceived flaws in competing products rather than highlighting the particular sweet-spots for their own. I implemented a master/worker pattern for order management in GigaSpaces a while back, and having recently spoken to the business director that still owns it, can attest that, as John says, we chose 'the right tool for the problem'. However, these days I'm working for a trading organisation, real-time events and parallel processing are high on our agenda, there are subtle (and sometimes not so subtle) differences in operational semantics of the front and back offices that will influence our choices. As we move toward product selection it's information that’s useful, not squabbling.

   Reply

7. Back to top

   Re: Paremus Infinflow

   Oct 16, 2007 6:02 AM by John Davies

   Richard, You're right and I'm sorry I didn't mention Infiniflow, there was a point when I thought, shall I make an exhaustive list of players at this point or just stick to my usual "group" of vendors, well you can see the direction I chose. You guys have given me a great demo of Infiniflow and Newton and it's definitely the most sexy grid demo I've seen to date. I know some of your customers but unlike the ones I talked about I've not had hands-on experience so felt unqualified to talk about what you do in detail. I hope this will change some time soon. -John-

   Reply

8. Back to top

   Re: Few comments

   Oct 16, 2007 6:26 AM by John Davies

   Nikita, I saw your presentation at JavaZone, GridGain is seriously cool but as per my comments above, I've not had a huge amount of hands on and I've not seen it in any of the banks I'm working in yet (YET) so I avoided it, perhaps wrongly. It's the simplicity of GridGain that makes it so powerful and I expect to see it playing a bigger part of Java-based grid in the coming months. I did make a conscious effort to avoid DataSynapse and Platform despite having used both in a large bank, one of whom currently funds one of the above. My reason for avoiding them was in this paper was simply because they're not Java specific and I was positioning this towards Java readers. I've now dug myself a small hole because this means I really should have mentioned GridGain, version II perhaps. I disagree with you about JavaSpaces being more data-grid than compute-grid but as you say it just depends on how you chose to use it, the point we are both making is that you can use it either way. -John-

   Reply

9. Back to top

   Olso?

   Oct 16, 2007 9:05 AM by Kjetil Paulsen

   "I caught up with John Davies while at JavaZone in Olso Denmark." ... JavaZone -> Oslo -> Norway :) ... Thanks for a great time guys.. Kirk, Cameron, John - great presentations, great fun..

   Reply

10. Back to top

    Re: Olso?

    Oct 16, 2007 11:42 AM by John Davies

    "I caught up with John Davies while at JavaZone in Olso Denmark." ... JavaZone -> Oslo -> Norway :)

    For the American readers, Oslo, Denmark and Norway are all in northern Europe. :-) I know what Kirk meant to say, we actually met up at JavaZone in Oslo, Norway and then the following week at JAOO in Arhus, Denmark. Last week we met in London and later this week we'll meet again, this time for some wine tasting in Tokaji, Hungry rather than Java. I'm sure the topic will come up though as I'll "stuck" be in a wine cave on Saturday while England play in the Rugby World Cup final. :-( -John-

    Reply

11. Back to top

    Re: Olso?

    Nov 11, 2007 3:39 AM by Peter Foldi

    Tokaji, Hungry -> Tokaj, Hungary :) Peter - from Hungary

    Reply

12. Back to top

    Re: Olso?

    Jan 23, 2008 10:08 AM by Cameron Purdy

    Peter, don't take it personally .. I was there with John and his drinking does impair his spelling ability ;-) Peace, Cameron Purdy Oracle Coherence: Data Grid for Java, .NET and C++

    Reply

• Older >