Terracotta and Continuations: Video Tech Brief


News: Terracotta and Continuations: Video Tech Brief

  1. Terracotta and Continuations: Video Tech Brief (9 messages)

    Geert Bevin gives a concise overview of how Terracotta clustering works and describes how easy it is to integrate it with your applications in a transparent way, unlike other clustering technologies that require programmer participation to make them work. "Terracotta's philosophy is to treat clustering the same way as garbage collection," he said. Geert also talks about his work in continuations and clustering, and gives a few examples of how continuations work, and why they're important to web developers. (Click here if you can't view the video.) Geert is a developer at Terracotta Inc., is the CEO and founder of Uwyn bvba/sprl and created the RIFE project which provides a full-stack Java Web application framework for quickly building maintainable applications. He started or contributed to open-source projects like Bla-bla List, OpenLaszlo, Drone, JavaPaste, Bamboo, Elephant, RelativeLayers, and Gentoo Linux. Geert is also an official Sun Java Champion. Watch other Tech Briefs

    Threaded Messages (9)

  2. Thanks a lot for posting this tech brief! If anyone wants a detailed explanation of clustered continuations with Terracotta, together with a simple example to try out, you can read the article that Jonas Bonér and I wrote: http://www.artima.com/lejava/articles/distributed_continuations.html Take care, Geert
  3. Good piece! I like the cliffhanger too: "And actually, I am..." ;-)
  4. "Terracotta's philosophy is to treat clustering the same way as garbage collection,"
    I think this whole premise is rather wrong. You can’t equate in-VM memory management (a.k.a garbage collection) with distributed cluster management. I have been saying that Terracota technology is more about “Look, ma, no hands!” type of technology rather than a necessity driven product. I bloged more about it at http://www.jroller.com/nivanov/entry/gridgain_vs_terracota_seeing_through Best, Nikita Ivanov GridGain - Grid Computing Made Simple
  5. Real world: I've been using Terracotta DSO(clustered a lucene RAMDirectory), was able to get a 3 node cluster running in no time ... it simply works the GUI to monitor the nodes is great also.
    “Look, ma, no hands!” type of technology
    nice bias post Nikita LOL, you should at least respond with some "real" advantages and disadvantages, not some rant quote... if at all the philosophy for Terrcotta would be "Keep it simple stupid" in regards to clustering.
  6. “KISS” is good. But coming off with another paraphrased sound bite I can say that “middleware should be as simple as possible BUT not simpler”. Besides the rant (which is perfectly natural coming from obviously opinionated and somewhat biased individual :-)) I was mostly replying to the fact that many confusing Terracotta technology with traditional grid computing when in reality it is not. I wish folks at Terracotta would improve their message and/or positioning (at least on the website) to make it clear what Terracotta technology is... Best, Nikita Ivanov. GridGain - Grid Computing Made Simple
  7. Come on[ Go to top ]

    Nikita, Agreed with prior poster, what is your issue? I have been arguing for the past 2 weeks internally that GridGain on top of Terracotta would be really cool, and totally doable by the looks of your SPI which seems really nice and well thought out. Your continued opposition to us just for the sake of opposition is not helping that cause, nor does it make you look very objective in the space.
  8. Re: Come on[ Go to top ]

    I have been arguing for the past 2 weeks internally that GridGain on top of Terracotta would be really cool
    I considered the same approach, but ultimately came to the conclusion that it's just too much of too new technology. Yes, it would be really 'cool', but i think, at least in our case, that the primary need it would had satisfied would had been developers' personal ambitions instead of real business needs. Terracotta makes it possible to write custom map/reduce, master/worker griddish stuff really easily, and using only terracotta keeps things a lot simpler compared to giving birth to terracotta/gridgain crossbreed bastard child.
  9. Re: Come on[ Go to top ]

    "a lot simpler compared to giving birth to terracotta/gridgain crossbreed bastard child" I believe one of the benefits of the GridGain solution was that it could be (it is) layered on-top of existing data-oriented grid solutions providing a consistent job/task oriented programming model that hide to some degree underlying deployment and processing distribution APIs which could also include the NULL API + XML + "align your class (root) design" approach. GridGain today allows an organization to move from a JGroups to GigaSpaces or Oracle Coherence depending on the budget and the degree of scalability and reliability required. Seems like a reasonable risk management practice to me. We recently spent time integrating our performance management solution and it has been significantly easier than other solutions because of extension points that have been clearly and cleanly delineated within the design and code. Kind regards, William Louth JXInsight Product Architect CTO, JINSPIRED "Java Performance Monitoring and Runtime Diagnostics for Java EE, SOA, and Grid Computing" http://www.jinspired.com
  10. Re: Come on[ Go to top ]

    Alright, I will drop my 2 cents :) I don't think Nikita has any opposition to Teracotta at all, but as usual, when anyone disagrees with some approach on TSS, there will be people who will get offended. If you look at Nikita's blog, he does suggest that a product like Terracotta is more low level than GridGain and potentially could be used to create a product like GridGain. However, I do think that criticism should be welcomed by Terracotta people. In my opinion (and again you may disagree) Terracotta is the kind of product that will take you for a ride but does not let you hold the steering wheel when you need it. For example, due to lack of API, I don't think Terracotta will allow you to do any pluggable custom node provisioning or job collision resolution or custom logical task splits or weighted splits, etc... What if, for example, I have a node that is available only during off-peak hours? Or what if one node is twice as powerful as others - can I give it more load? Or, moreover, can I choose exactly what kind of load to give to that node? For the same no-API reason, Terracotta is not pluggable at all and does not let you blend into existing environments. For example, if I am using JBoss server, I can't utilize existing JBoss underlying communication protocol for Terracotta or provide a custom load balancing. I do think that Terracotta is a nice clustering product, but it seems that due to the chosen no-API approach, it does have many limitations. Best, Dmitriy GridGain - Grid Computing Made Simple