February 14th, 2005
EMIC extends open source clustering to Tomcat, JBOSS, and JOnAS
At this week’s LinuxWorld in Boston, Emic Networks — provider of fault tolerance and load balancing solutions for open source server software such Linux, Apache, MySQL and PHP (collectively known as the LAMP stack) — is extending the coverage of its clustering umbrella to open source-based Java-based application servers. In ZDNet’s 10th podcast audio interview (download the MP3, or learn how to have them automatically downloaded while you’re sleeping), I discuss the announcement with Emic’s vice president of product management and marketing Donna Jeker.
At the end of the interview, Jeker initiated a comparison of Emic’s technology to that of Oracle’s and made it seem as though you get the benefits of Oracle’s clustering solutions for a fraction of the cost. Ultimately, yes, both solutions deliver a degree of fault tolerance, scalability, and manageability. But the approaches to providing those services and the degree to which the solutions can power massive, mission-critical applications are so different that I can’t help but wonder if that’s like saying that a Kia offers equal protection to that of a Hummer because both have airbags. While Emic’s solutions may very well be worth the investment for what you get (as are many low-to-midrange clustering solutions), just remember you get what you pay for. I’m not sure making the comparison to Oracle is a good idea.
Here are some of the questions and edited excerpts from her answers:
What is Emic announcing at LinuxWorld?
Describe a typical configuration and the beenfits.
Is there a limitation to the number of systems that can be used to assemble a cluster?
If there’s a failure in the middle of a transaction, regardless of whether the transaction is a J2EE transaction or a Perl script transcation, Python, whatever, does the cluster just continue to run, business as usual? Clients don’t see any sort of interruption? Or must someone like a sys admin or app manager go in and study some sort of roll-back log and reassemble the transaction to make sure that everything was running the way it was before things went bad?
So to be clear, the type of failure that you protect against is what?
Do the the transactions restart themselves or does the user actually have to physically recognize that the transaction needs to be restarted and take action?
Some of the other questions that I asked:
- You’re describing the type of failure that happens when you’re on a web site, for example, an e-commerce transaction. What about with J2EE, which is what you’re announcing today. With J2EE, a lot of the transactions, a lot of the workflow, take place behind the scenes….. How does it take the same action that an end user might have to take if they’re sitting in front of their browser and they realize they have to restart?
- What sort of interconnect is required between the different systems in the cluster?
- Do you certify your software to work well and be supported by specific J2EE servers? Does it have to be certified by Sun to be J2EE compliant?
- How are Emic’s products priced (including the newly announced product)?
- Oracle’s clustering uses a shared-everything approach; it gets faster and more resilient as you add more systems. Do you use a shared-everything approach?
