Discussions

News: Major update to WebSphere XD DataGrid shipped

  1. Major update to WebSphere XD DataGrid shipped (18 messages)

    IBM WebSphere XD DataGrid V6.1 which includes the ObjectGrid data grid middleware has just shipped a major update to the ObjectGrid function called iFix 3. ObjectGrid provides an embeddable distributed memory platform for implementing network attached memory, HTTP Session management, complex event processing, next generation scalable OLTP databases or Extreme Transaction Processing (XTP) style applications. It provides a virtualized partitioned, replicated data store. Data in the store can be accessed using a data style (relational/map) or in a messaging style (FIFO consumers) simultaneously. Data can be fully indexed and retrieved using its key or using queries on any indexed attributes in the store. The data store can be hosted in the same cluster as your application or as a separately hosted cluster. As the cluster grows and shrinks, ObjectGrid automatically scales out and in to best distributed the data store across the current set of JVMs hosting the data store. The collective memory of all JVMs hosting the store is all that limits the amount of data that can be stored within the data store. It provides Map based APIs to access the data or an EntityManager style API as well as a stored procedure like API (DataGrid APIs). It's very lightweight and scales from a pair of JVMs to thousands distributed across multiple data centers. It integrates fully with WebSphere ND V6.0.2 and ND V6.1 and can be used without WebSphere using just J2SE 1.4.2/Java 5 with your favorite application container. When packaged for J2SE it's a couple of jars under 30MB in size. Just add to classpath, edit some xml and go. The new function enables several major features including support for data center replication, treating data in Maps as queues and Spring integration. More can found on the ifix 3 release notes. The Spring integration enables the use of Spring native transaction support and its container managed transaction facilities. It can also use Spring Bean Factories to obtain references to ObjectGrid extensions which allows them to be easily wired into a Spring based ObjectGrid application. The fully functional J2SE evaluation version has already been updated to the latest level and now runs for 8 hours at a time. It is available for free download here. The evaluation version just requires a Java IDE and a JVM to test. Customers can download the ifix here.

    Threaded Messages (18)

  2. Congratulations Billy and team! The Spring integration looks very interesting. I think it could be made more transparent using a custom scope for objects managed by ObjectGrid. Have you considered this? Rgds Rod - SpringSource
  3. I have not had a chance to evaluate it yet, but I have several questions: 1. Does it work outside of WebSphere? 2. How does it compare to existing Data Grid products (Oracle Coherence, GigaSpaces, JBoss Cache, etc...). I would be interested to find out the comparison from usability and performance point of view. 3. What does "virtualized partitioned" mean? Is it the same concept as partitioned cache in Oracle Coherence? We at GridGain team would definitely be interested in integrating with it to complement our Compute Grid capabilities. Best, Dmitriy Setrakyan GridGain - Grid Computing Made Simple
  4. Yep 1. It works without WebSphere. Put objectgrid.jar in your class path and thats it. If you want byte code rewriting then you need cglib.jar, spring support ogspring.jar and stream query support then the ogquery.jar. 2. There is no outright winner in this space unfortunately :( All products are getting closer and it depends on what you want. ObjectGrid has probably the best replication technology with the new zone support and the EntityManager API offers a very nice programming model for writing CTS (contrained tree schema) style highly parallel applications. We are also very easy to embed in your existing application server (no matter who makes it) or just run in your j2se process. As far as open source caches I don't think there is anything that competes with the commercial brethren quite yet. This isn't to say offerings like ehcache, terracota etc don't have their use cases and do them well but they don't have the same design goals as the commercial offerings if thats a good way to put it. 3. Virtualized just means that association of a partition primary or replica (we call them primary shards and replica shards) to JVMs is dynamic. We look at shards that need placement and then place them if possible according to the placement rules defined by the customer. If new JVMs are started then 1/N of the existing shards will be moved to the new JVM to balance the distribution. If a JVM fails then we first handle failover and promote surviving replicas to primaries etc and then we handle compensation, i.e. add more replicas to bring the level of redundancy back to the level asked for by the customer. We describe the data store as virtualized on the current set of JVMs that are acting as shard containers. Email me if you want more information. I'm easy to find.
  5. Interesting. I will take a look. One of the "problems" we trying to solve is how to factor away the "relationalness" and "databaseness" from all of the users including analysts. Otherwise people spend all their time trying to get to the point of trying to find answers with the "data" instead of just finding answeres. Application ("OLTP") users have good tools to do this. And with tools like BIRT and Crystal Reports we can do reports agains the objects. But when we try to use the same object model with OLAP (both offline and online), it is difficult because these are still tightly coupled to databases and data structures. We would like everyone to use the same objects (data/relationships/logic combined) and be able to use an JOQL type language. And we want to use grids or object spaces because they solve the performance issue, but introduce advanced querying issues. Billy, can you provide links to the DataGrid querying APIs? I will be looking on the website on my own too.
  6. Re: Major update to WebSphere XD DataGrid shipped[ Go to top ]

    Interesting. I will take a look.

    One of the "problems" we trying to solve is how to factor away the "relationalness" and "databaseness" from all of the users including analysts. Otherwise people spend all their time trying to get to the point of trying to find answers with the "data" instead of just finding answeres.

    Application ("OLTP") users have good tools to do this. And with tools like BIRT and Crystal Reports we can do reports agains the objects. But when we try to use the same object model with OLAP (both offline and online), it is difficult because these are still tightly coupled to databases and data structures. We would like everyone to use the same objects (data/relationships/logic combined) and be able to use an JOQL type language. And we want to use grids or object spaces because they solve the performance issue, but introduce advanced querying issues.


    Billy, can you provide links to the DataGrid querying APIs? I will be looking on the website on my own too.
    I think I found my answer to the querying question, which is "It does". I would like to know if any knows if anyone is working on solving the other issue.
  7. As far as open source caches I don't think there is anything that competes with the commercial brethren quite yet. This isn't to say offerings like ehcache, terracota etc don't have their use cases and do them well but they don't have the same design goals as the commercial offerings if thats a good way to put it.
    I would agree with you in general that most of the open source caches are not competing yet with commercial offerings, but that is not to say that all of them are not there yet. JBossCache, for example, significantly improved performance with addition of Buddy Replication and scales quite well now. With addition of data partitioning (which should be coming soon) they should get pretty close feature-wise to many commercial cache products. Best, Dmitriy Setrakyan GridGain - Grid Computing Made Simple
  8. 8 Hour Trial[ Go to top ]

    Too bad the trial lasts only 8 hrs.
  9. Re: 8 Hour Trial[ Go to top ]

    Too bad the trial lasts only 8 hrs.
    Just running it does. If you need a bit longer, just schedule a task that resets your PC's clock.
  10. Re: 8 Hour Trial[ Go to top ]

    You can use the ObjectGrid trial as long as you want I think but you can only run it for 8 hours at a time which is plenty for class room situations, kicking around the tires, benchmarking etc.
  11. IBM WebSphere XD DataGrid V6.1 which includes the ObjectGrid data grid middleware ..
    Hi Billy - Congrats on the release .. looks like you've been busy ;-) Peace, Cameron Purdy Oracle Coherence: Data Grid for Java and .NET
  12. XA Transactions..[ Go to top ]

    Looking at the release notes, it doesn't look like it supports XA Transactions, and only supports the WebSphere App Server Transaction Manager. Any plans for a JCA Resource Adapter that will work with any App Server and has support for XA? (Like Coherence?). Product looks great, smells just like Oracle Coherence.
  13. Re: XA Transactions..[ Go to top ]

    "Smells like coherence". I'm sorry, I just can't make that sound good :) We don't support 2PC because of performance concerns. We do support running as a last one phase resource on a JTA transaction. The provided plugin to enable this with WAS is around 30 lines of code using public APIs and it should be easy to write one for any J2EE server/JTA provider that allows the the plugin to detect if a JTA transaction is already active and if so, the plugin registers a callback for commit/rollback notifications. The plugin just does implicit begin/commit/rollback based on these events. If and when we add 2PC then JCA would be the way we'd do it. The OpenJPA support consists of a 2nd level cache plugin which attaches to a grid of JVMs with or without a near cache enabled depending on what's needed. It caches entities and queries. It also takes advantage of OpenJPAs new more efficient cache plugin interface developed with ObjectGrid in mind but clearly could be leveraged by the anyone. We will also have a JPALoader that makes it very easy to use JPA Entities as Map key/value pairs. This makes backing a Map with a database table very easy. I have the JPA Loader packaged as a sample right now (it's a great example on writing loaders) but we'll likely ship binaries with the product shortly.
  14. ObjectGrid with OpenJPA?[ Go to top ]

    Billy, I'm looking at your ObjectGrid Overview presentation. I don't know when the version of the presentation I have was created. It mentions that ObjectGrid will have support for OpenJPA "later this year." I know ObjectGrid already supports an EntityManager-like API. When do you expect the OpenJPA support will be available?
  15. Billy congrats for the release - i look forward to see you in the next event. Nati S. GigaSpaces Write Once Scale Anywhere
  16. Cangrats guys[ Go to top ]

    Looks like I've got a lot of reading and code to play with over Christmas, well done Billy and team, I look forward to taking a look. Any chance you guys are presenting at TSS this year? -John-
  17. Re: Cangrats guys[ Go to top ]

    If we get an invite, yep.
  18. I have been looking at Cacheonix for the last couple of weeks. I have managed to get very early access to their cache manager and it does and we want. We will most likely go with Cacheonix, but I would like to look at ObjectGrid anyway. Does it provide programmatic cache configuration in addition to XML files? Raymundo
  19. Darn,
    ... *what* we want...
    Raymundo