Re: partitioned? distributed? parallel? replicated?

Brant Cooper wrote in message <36BF4E5B.B805A555_at_tumbleweed.com>...

>Thanks for your comments. The answer to why is scalability. Leaving
>aside redundancy and fail-over, what my customers want is the ability to
>add hardware to existing systems over time.

Adding hardware as oppose to upgrading means in my mind Oracle Parallel Server. But even that does not mean you can simply throw more hardware together. For example, with OPS you need a Distributed Lock Manager (DLM). The DLM is provided by the operating system. That of course ties you to the operating system and even the hardware.

Also, customer expectations are often wrong. Throwing more hardware at a system does not solve performance problems or magically increase the OLTP rate. And often it does not even make financial sense. I mean, they won't go and add another V6 engine to their custom Chevy to make it go faster would they? But many times, customers expect this from the computer hardware world. And managing expectations can be a full time and a pain in the butt job. :-)

To upgrade the hardware or to scale by adding more hardware? That is the question. And now I'm all out of quotes for the day. The answer is not simple. Clusters and massive parallel processes (MPPs) are COMPLEX! Period. With this complexity comes additional risk. Specialised skills. And a host of other assorted fortune cookies. This can all be managed with the right people and the "right" budget. Which you very seldom get in my experience.

I like the car analagy so let me use it again. These high-end systems are like F1 Grand Prix cars. They need a lot of attention, a good pit crew and good driver. And they need to be run on proper race tracks and not on dirt tracks. They are made for racing, and not pulling U-Haul trailers or ploughing. If a customer knows this, accepts this and have realistic expectations then GREAT!! Please introduce me to your customer because it's not often that you find these type of levelheaded business people down here in Africa. :-)

>My customers would
>rather start with an E450 and add another one if necessary, than start
>with an E4500. Are the complexities and costs justified. You betcha!

With enough money... Reminds me of what some of the fighter jocks said about the F-4 Phantom. With enough power even a brick can fly. With the right budget, risks and complexities can be managed - no arguments from me there. The question though is, will you have a good night sleep carrying the can for this.. ;-)

My 1'c worth. (government taxes too the other cent)

regards,
Billy Received on Tue Feb 09 1999 - 01:10:47 CST