The newest version of Sybase's database management system -- Adaptive Server Enterprise 15 -- has entered the second phase of beta testing and is due for commercial release this fall. SearchEnterpriseLinux.com contacted Tom Traubitz, Sybase's senior marketing manager for the Adaptive Server Enterprise product line, to find out what Linux users can expect from the new release in terms of features and functionality. Here are some excerpts from that discussion:
What are the major areas of focus for the upcoming release of Adaptive Server Enterprise?
Tom Traubitz: The major focus of this new release is about solving when your data is growing enormously. There is a lot of technology such as partitions, and more self-management technologies in there. [We've included technology for self-managing] internal statistics pages. One of the things that in optimization that is very important is to keep statistics about what the data look like. Many systems, such as our original system, required you to manually update those statistics if the data changed a lot. Now that is all automated. There are also a lot of things we've done around query optimization and messaging to improve velocity.
What have you done specifically in terms of messaging?
Traubitz: The latest version features a new messaging component technology primarily supporting Java Messaging System, but we're also extending that to MQSeries. That will be in the next release. But on Linux I think you're going to find most people use JMS right now.
What numbers do you have regarding Sybase's performance on Linux?
Traubitz: Currently, on a four-processor Linux box, I think we're getting between 150,000 and 160,000 transactions per minute. Analysts tell us that that covers 70% to 85% of existing database management problems.
How does Sybase's licensing model work?
Traubitz: We have multiple licensing models. One model is server and seats. You pay a price for the existence of the box and then you pay a per user charge. That starts at $1495 for five users. And then we also have CPU-based pricing where essentially you just pay by the number of cores in the box. That is $4995 per core. We also have an express edition that is limited to users with five gigabytes of data under management and one central processing unit (CPU) right now. That express edition is free for deployment and is primarily to support the Linux community.
Sybase executives often talk of an 'operational gap' in technology. Can you explain what this concept?
Traubitz: The operational gap is really this: The machinery's ability to handle bigger amounts of data has certainly [improved.] But our ability to really manage that effectively - being able to back it up, being able to verify its consistency, being able to handle the volume of change that probably is associated with that amount of data -- has not grown as quickly.
Traubitz: Partially it's because of people issues. When you get bigger and bigger databases, you're putting more demand on people. And most of our clients can't afford to have the number of people managing their systems grow linearly with the amount of data they're managing.
What technology does Sybase hope will address this gap?
Traubitz: A lot of the technology is around self management. Homeostasis in biology is the tendency of an organism to maintain equilibrium. What we're trying to do is draw a parallel to that in technology. When the system encounters things in the environment such as, 'I'm getting a little low on memory,' or, 'I need to adjust how I'm using memory,' - it should not have to call for a DBA to come in and retune it. So, a lot of self-tuning is in the underlying technology.
What is the specific underlying technology that makes this self-tuning possible?
Traubitz: Sybase has some very interesting technologies in our database which make up our Monitoring Framework. What it does is it maintains a lot of in-memory counters of different attributes of the server, such as where locks are being taken out, what tables are being hit particularly hard by different types of questions.
You mentioned that Adaptive Server Enterprise is designed to be looked at as a single process by the operating system. Why?
Traubitz: Sybase tends to look like a single process to the operating system. And inside of it we then assign for every user a thread, and we do our own switching between the threads. We try not to have the operating system make those judgments for us. The reason why is that the operating system tends to make those judgments on kind of a grand operating system scale. But the operating system is not the database, so it doesn't know what is going on in the locks, or who is using what resources. So, what will happen when the operating system makes these decisions is tit will sometimes makes bad choices. It will give time to somebody who is waiting for a resource. It may contact switch to a thread that can't do anything right now. Because we're the database, we know all those things and can time slice more efficiently.