Home

Database 11g - change the rule of server hardware planning

Submitted by iniunet on Tue, 07/07/2009 - 21:45

It has been common sense that hareware is getting cheaper by every year, Nobody is surprised that today's server is 10 times faster than a server of 10 years ago, yet only ten percent of the previous price.
I used to see many customer buying large server even before the project going-live. cause everything must be ready before going-live. NO change will be allowed to made to the production hardware after going-live except that something wrong or too much deviation from the previous estimation. The server processing power, in this case, must be pre-calculated and prepared in advanced.
Some of these systems, might only have hundreds of transaction at the beginning that could be covered easily by a small PC server.
Hardware planning dilemma : Question and Answer

  1. So what is the use of a super large server processing power?

    The answer: It is prepared in advanced that after a few years, there might be millions of transactions!

  2. Why buying tommorrow 's power using today 's money when this processing power might be only ten percent of the price as time goes by!

    The answer: Change is agonizing, changing the production system is risky and unacceptable, we 'd rather spend extra one million!

    But we 're in the crisis right now, and we do not have extra one thousand, not to mention one million.

  3. The pre-calculation of processing power might not be accurate, If the processing power requirement never be exceeded at the time we retire the system, just Thanks God! But what if we under-estimate the requirement, now we will have to do some change.

    Change might be inevitable at last, It is only the price of the change that matter.

Hardware planning in 11g : we will pay as we grow!
To ease the change of infrastruture seems to be a long term of objective of Oracle database.
Actually RAC (real applcation cluster) is pioneering in the 'pay as we grow' strategy. All we have to do is to add one or more node to the database to provide nearly unlimited scalability as the requirement grow.
By introducing the Oracle VM under the enterprise linux, The change will be more easy , even the single node database could be easily relocated to an more powerful server with the righ processing power.
It is now possible to patch the database online by relocating the instance to another server and applying the patch and swithing back to the patched node.