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 (72045)	2/20/2002 4:54:26 PM
From: Tony Viola	Read Replies (1) \| Respond to of 275872

Fyo, The problem with that is that clusters are just simply not as stable. How do you mean? Do you mean reliability-wise? If so, clusters can be put together with no single point of failure (SPOF), and failover, and can be made quite reliable. Tony

To: fyodor_ who wrote (72045)	2/20/2002 6:00:27 PM
From: eCo	Read Replies (1) \| Respond to of 275872

fyo:The problem with that is that clusters are just simply not as stable. Link please. :) According to the Linux parallel processing 'HowTo', one of the upsides to a cluster is: "The fact that replacing a "bad machine" within a cluster is trivial compared to fixing a partly faulty SMP yields much higher availability for carefully designed cluster configurations." However, the downsides are: "With a few exceptions, network hardware is not designed for parallel processing. Typically latency is very high and bandwidth relatively low compared to SMP and attached processors." And also: "There is very little software support for treating a cluster as a single system." linuxdoc.org remember the Enterprise Customer survey... Yeah, I saw that. But that list is based on a paradigm of non-replaceable boxes. If you've got $2-3k machines in there instead of $10-15K boxes, swapping a unit out becomes more cost-effective than putting more resources in "reliability" and "quality of service". All IMHO, of course. eCo

To: fyodor_ who wrote (72045)	2/20/2002 9:07:43 PM
From: Dan3	Respond to of 275872

Re: The problem with that is that clusters are just simply not as stable. I really think that you are mistaken. Whenever high reliability is needed, clusters are the first choice.

To: fyodor_ who wrote (72045)	2/20/2002 9:15:46 PM
From: pgerassi	Read Replies (1) \| Respond to of 275872

Dear Fyo: I think you have it backwards. More and more server backend are becoming clusters because they solve those higher priority reasons better. Lets take them one at a time: 1. Reliability Clusters are more reliable from unavailability and total available average processing power standpoints. And you also have the option of distributing some nodes to a remote location to prevent WTC like disasters. 2. Quality of Service and Support Service and support on 10 similar simple machines is cheaper than on one very complex machine 10 times larger. 3. Performance MMPs (always cluster based) are still the top performing machines in any SC class listing. 4. Total Cost of Ownership 10 simple processing nodes are cheaper than 1 10 processor system. You save on initial purchase, you can scale when you need to without throwing the old nodes out, you save on service, you save on support and finally you save on disposal. 5. Compatibility with existing systems This is actually the same, unless you are clustering existing systems and maybe adding more new nodes. Then the advantage goes 100% to clustering. 6. Scalability This is well known that clusters can scale far more than any SMP box ever designed. Scaling also can be done in smaller amounts, as little as just one new node at a time. 7. Reputation of Vendor Certainly IBM has a better reputation than DELL or your local SD shop. But, in some cases, the local SD shop is willing to bend over backwards to accomodate your needs and give you that very desirable hand holding that many like so much and do not get from the big tier one OEMs. 8. Ease of Administration Well written utilities are available to both types with SMP having more history. 9. Initial Purchase Price Clustering wins this hands down even when you are getting more when you cluster over SMP. 10. Interoperability in Multiplatform Environments This also depends mainly on your application writers (or who you pay to do this). I certainly had no problems interfacing systems from widely disparate vendors and types (from embedded CPUs to mainframes). In looking over the results, clustering appears to have the advantage, which is certainly not lost on IBM, Sun, Compaq and many others. Pete