Middle-tier diet…

 

Tom Kyte has written a nice piece on How to scale.. Suffice to say, I agree entirely with him, as previous posts on this blog will testify, but I wanted to look at this from a different, maybe cynical, angle…

For a hardware manufacturer, doing more work in the middle tier is a dream come true. Sure, J2EE can scale well, mostly because it has to! You need a server room like a Google data-center to run the Pet Store demo application. 🙂 Encouraging companies to invest in middle-tier processing is a massive win for the hardware manufacturers, especially the Intel/AMD based guys.

From the software licensing perspective, this is a massive win also. Scale out in your middle tier and pay us lots of money for licenses. No wonder Oracle are laughing all the way to the bank. In my company, we probably pay more for our application server licenses than we do for our database licenses, yet I’ve worked on far more complicated projects that manage with a single Apache server. We’ve been encouraged to believe that if it doesn’t need 10 clustered application servers, it’s not worth having, when the reality is, it needs 10 clustered application servers because the middle tier stinks!

We seem to be a society that relies on quick fixes. If we want to lose weight we go on a fad diet, get our stomachs stapled or have cosmetic surgery. Middle-tier processing and scaling is the cosmetic surgery of the IT industry. It doesn’t matter about fixing the root cause, just brush over the surface and everybody’s happy… Right?

Wrong! Put your fat-ass application on a middle-tier diet!

Cheers

Tim…

Author: Tim...

DBA, Developer, Author, Trainer.

6 thoughts on “Middle-tier diet…”

  1. When I first encountered Siebel, I was horrified to learn it didn’t support native referential integrity constraints, table partitioning, parallel query, IOT’s, parallel DML, CBO and all those other Oracle features.

    I was convinced all of this would significantly hinder the scalability of the middle tier.

    4 years on, it turns out it doesn’t. Granted, there is *potential* for better, tighter integration with Oracle DBMS and who knows, given recent events, that may well come to pass in the next few months.

    However, I have Siebel running on a 4 year old laptop with 1GB of memory (Web server, Application tier and Oracle 10g database) and a Linux VM. I tell customers this who moan about Siebel’s massive hardware requirements and memory footprint. [Hint: look carefully at your customisations and scripting before you buy yet another cluster]

    ‘Lowest common denominator’ database support from Siebel and other vendors may well infuriate the purists and Oracle evangelists but makes supporting other databases (DB2, SQL Server) a lot, lot easier.

    Surprisingly, database independence is important to application vendors for commercial reasons (license sales). There are still of lot of Big Blue customers out there, typically in the big corporations who hold big IT budgets.

  2. j2ee is the biggest rip-off in the history of IT.
    there is not a single claim that j2ee makes that has ever been sustained and proven in real life. not one.

    but it sells hardware as if it was hot cakes. and OS licences. and app server licences. and “seats”. and a lot of other things.

    as such, there isn’t one marketeer of one of the above that doesn’t like it.

  3. 4 years on, it turns out it doesn’t. Granted, there is *potential* for better, tighter integration with Oracle DBMS and who knows, given recent events, that may well come to pass in the next few months.

    However, I have Siebel running on a 4 year old laptop with 1GB of memory (Web server, Application tier and Oracle 10g database) and a Linux VM. I tell customers this who moan about Siebel’s massive hardware requirements and memory footprint. [Hint: look carefully at your customisations and scripting before you buy yet another cluster

  4. Pretty impressive. Never tried middle-tier diet yet.

    I just tried checking other blogs regarding this middle-tier diet. Most of the people have at most 70% success rate in applying such. I am thinking if it would be risky to my part as a starter or not.Nevertheless, I’ll just try this out soon.

    Thanks sharing though.

Comments are closed.