Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: fyodor_ who wrote (72073)	2/20/2002 10:11:17 PM
From: combjelly	Read Replies (1) \| Respond to of 275872

"There are very few applications available that will work on a cluster out of the box." Have you looked at MOSIX? Or are you stuck with Microsoft?

To: fyodor_ who wrote (72073)	2/20/2002 10:48:45 PM
From: pgerassi	Respond to of 275872

Dear Fyo: RE: "Uhmmm... NO! Compatibility is zero. You need to start from scratch, doing everything yourself. There are very few applications available that will work on a cluster out of the box." BS!!! I have done this with clusters of telephone switching systems. These still do all of the things that they have done in the past and they have always been clustered. You do not get the uptimes they have with single massive mainframes even doubles and triples. And you can't down a portion of an SMP box sitting on a mountaintop where it takes a all day helicopter ride in good weather to fix. You can with a cluster. In fact you leave certain nodes up and half used to have spare capacity in case you need to down a few nodes and do that copter trip when its convenient not, absolutely required. Are there problems with clustering? Yes, but they are being rapidly reduced. And inside most of those IBM mainframe and their replacements are built from a clustering standpoint. This includes all of their IA-64 systems, Power4 system and many others. You need to go into the underlying architecture documentation on their newest systems to see this. Many other company's n-way servers are built with similar clustering within. They also are serviced the same way. Perhaps those mainframes you talk of are really clusters internally (probably are) and you did not know it. That allows that essential quick to repair and the ability to lose a node and keep on ticking. As to task distribution, whole task are easy to distribute, just find the active node with the lowest use. It does not need to be perfect, just good enough (the algorithms are quite simple and do a surprisingly decent job). If you perform using the latest RDBMSes, you can get a distributed cluster based database up and running and your applications merely become tasks distributed to the nodes (usually different that those holding the raw data but, they do not need to be) and they in turn talk to the RDBMS as to what they want done. The RDBMS takes control of what is stored where and what to tell the underlying nodes with the raw data. The programmer never knows that he/she is on a cluster and to them looks like any large server of old. Thus applications are compatible (this is how Oracle promotes their latest and greatest) and your arguments go right out the door (isn't layering or abstraction wonderful?). Again IBM, Sun, Compaq (DEC), and many others use clustering methods at many different levels. DEC started the use of clusters in businesses and has always been a strong selling point in their favor. Networking people have been trained in the use of clustering technology and use it all of the time. Removing SPOF is quite a bit easier in a large network where all leafs have more than one way to communicate to any other. It is the same with clusters. Its possible to make a SMP box without a SPOF but, it is very difficult without using networking and thus, clustering techniques. Better check on your server documentation, you might find you are already clustering (just within a box rather than many boxes). It makes your other arguments moot. Pete