Discussions

News: Oracle challenges BEA with $1,000,000 performance contest

  1. Oracle has extended its million dollar performance challenge to include BEA's software, claiming that the company's database and e-commerce software will run Web sites three times faster than rival offerings from BEA Systems (or IBM and Microsoft), or Oracle will pay customers $1 million. A closer look at the fine print reveals the true nature of this shameless marketing campaign.

    So far, nobody had attempted to take on the $1,000,000 challenge since it was issued in October. Since Oracle's recent re-announcement of the contest to include BEA's software, BEA has offered free consulting to BEA customers to help them convert their applications to run Oracle and win the cool million. :)

    This challenge now technically applies to TheServerSide.com as well, since we are running BEA Weblogic. Although a million dollars would be nice, the Oracle challenge has been setup to make attempting extremely difficult, and winning even harder. A recent research note from the Gartner Group reveals the problems we would have to deal with if we took this challenge:

    1) According to Gartner, if we were to accept the challenge then our actual production box would have to be used to measure the performance difference, meaning that we would have to completely convert our runtime environment and risk introducing bugs onto TheServerSide, or even worse, going down for a long period of time during the conversion process. This is unacceptable to any real production system.

    2) The contest allows Oracle up to 90 days to tune our site to be 3X faster than it is currently running. This means that TheServerSide would be running in a potentially unstable state for up to 3 months, with people from Oracle poking around bringing the server up and down all the time.

    3) The conversion to Oracle is at our own expense, and the contest guidelines force us to use Oracle's consultants for any extra help we may need.

    In short, this contest is brilliant marketing on Oracle's part, but thats about all it is: marketing. :)

    Read the article on CNet.
    Read the Oracle Press Release.
    Read the Oracle Contest Rules page.
  2. You are absolutely right. This gimmick is an attempt to grab some new customers, rather than to convert existing WebLogic and WebSphere customers because that is almost impossible. At this moment, Oracle is losing badly in the Java app server space and I am not sure it can gain ground in the near future. If Oracle has any chance of picking up customers, the Oracle9iAS Cache functionality better deliver. I guess people are still skeptical about how good this thing is. Has anybody in this group used the iCache feature in Oracle9iAS?
  3. I havn't but the the popular, open source Resin Servlet Engine had a similar feature (caching of dynamic page output) a full year before Oracle. :)

    Floyd

    Check out Resin.
  4. There is more to it than the iAS Page cache. The new 9i oracle client can cache querable snapshots on the client. You can specify a refresh policy. Obviously, this is only good for read only data but it has a lot of potential for gains.

    Yes, containers can cache entity beans but what they can't do is run queries against the container cache, they usually run all (non primary key) queries against the database. If (and maybe it's a big if) you can take advantage of these querable cached tables then potentially you can get a big gain.

    The performance hype regarding java inside the VM is still over hyped. We've seen x 10 performance loss when compared to native stored procedures when you mostly just do data access in the proc. We saw this on all 3 major databases supporting Java. Oracle claim to have further tuned in 9i but I haven't test it yet. But, if you can do more calculation than data access then it swings the other way (x 10 improvement) over native stored procedures.

    If Oracle can sort out the internal jdbc performance then Oracles embedded EJB server combined with Oracle Parallel server is a very performant solution. You don't need to cache as your container is co-resident with the Oracle block cache, you can run several Oracle instances/ejb servers against a single database image and Oracles takes care of cache coherency. Of course, if the container doesn't cache (and it shouldn't) then you pay a big object construction cost but its worth testing in any case.

    Plus, add in their JMS implementation which you also get and it's transactional and fault tolerant. The JMS lets you send messages transactionally between clusters (something some well known EJB servers are still incapable of).

    I don't want to sound like an Oracle head, but once you know their game and techincal architecture, there is a lot of potential there and it is quite advanced relative to the competition. No, it's not the latest and greatest EJB spec but the basic functionality needed to get on with it is there.

    People say Oracle is expensive but add up how much WebLogic/WebSphere/EAServer on x boxes, Oracle on at least 2 boxes plus a real JMS solution on all boxes would cost and you won't think it's so expensive any more.

    As for the conditions, well you didn't think they would make it easy to lose, did you? I don't think Larry wears a red suit this time of year....

    Anyway, off for hols so happy christmas everyone....
    Billy
  5. I agree with Billy. Don't underestimate Oracle's experience with creating high performance systems. They've been out of the game for a while, but they do look to be back in it with 9i.

    We seem to be focusing a little too much on the negatives and not on whether iAS really *IS* faster than WL 5.1 or 6.0. Based on my experience, WebLogic is a "decent" server, but they're not the most technically advanced server (GemStone, i-Planet, Persistence PowerTier are arguably more advanced and faster, or WebObjects for a non-J2EE server).

  6. Are you attributing WebLogic's dominant market position purely on BEA's superior marketing execution? Oracle has a decent marketing effort but how come it cannot sell iAS as well as BEA sells WebLogic? WebLogic may not be one of the fastest app servers around but it definitely is easy to set up and deploy apps against (who knows how many more steps you have to take to run iAS?), in addition to its edge in time to market over all other app servers.