“An upgrade to Sun’s enterprise server clustering technology is set to arrive next week alongside the company’s much-anticipated new Solaris 9 operating environment. Backwards-compatible to Solaris 8, Sun Cluster 3.0 will offer features such as improved ease of management, enhanced dynamic configuration, and support for Oracle9i RAC (Real Application Clusters), according to Jim Sangster, the group manager for Sun’s cluster product line, in Palo Alto, Calif.” Read the story at InfoWorld.
Solaris 9 to Boost Clustering
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
The only thing Sun needs to do now is slash the prices on their hardware. Even with Sun clustering, it is cheaper to build a x86 cluster and have about the same performance. Yes, Sun’s hw is much more reliable, but when you have a huge amount of x86 nodes, reliabilty is not a problem.
wait wait wait – let me get this straight – you’re saing that an x86 cluster and a sparc cluster are comparable in performance. Would you happen to have some refferances to back that up? Personal experiance perhaps? Thought so. And what kind of clustering are you reffering to? HA or (implied by your e-mail) high-performance? An what kind of hi-perf cluster are you reffering to? Which configuration?
Dumb-Ass
Here’s a link for you. Buy the book, read it. Come back when you’ve gotten a clue.
http://www.amazon.com/exec/obidos/ASIN/0130084581/qid=1021480598/sr…
“The only thing Sun needs to do now is slash the prices on their hardware. Even with Sun clustering, it is cheaper to build a x86 cluster and have about the same performance. Yes, Sun’s hw is much more reliable, but when you have a huge amount of x86 nodes, reliabilty is not a problem.”
Yeah.. An x86 cluster can really compete with a Sparc cluster. There is no comparision at all. Why do you think Google has such a nightmare on their hands as far as hardware? Because they built their system on x86 hardware. Last time I checked, they had over 9,000 boxes running Google. All I know is I am glad I am not the one maintaining that nightmare. Had they gone with a more scalable archetecture like Sparc, they would probably have a much more managable system today. For example, a single E10K can scale up to 106 processors. And the UltraSparc III processors (Cheetah) will run circles around x86.
As far as cost, has made a lot of progress in some areas. For example, a SunBlade 100 costs less than the average PC does these days. My next dekstop workstation is going to be a SunBlade because I have always felt Sparc technology was superior to Intel and I an finally afford to buy one.
You’re just buying into the hardware conspiracy propagated by Sun, IBM and other big iron people. It’s just a big money grab by these companies. A couple of standard workstations put together via ethernet can do the exact same thing as all those big, fancy expensive servers. And I can build each those workstations for less than $500 with hardware I can buy at my favorite local computer store. They can put the difference between the Sun hardware and my cluster into my salary, and I’ll be rich rich rich!
Oh, did I mention I was being sarcastic.
>Dumb-Ass
speak for yourself
>they built their system on x86 hardware
I’m sure they had a reason to do it, or is google incompetent?
“I’m sure they had a reason to do it, or is google incompetent?”
They fell for the “free, open source OS will save you time and money” lines.
Saving money upfront does not always equate to a free and easy ride down the line.
“I’m sure they had a reason to do it, or is google incompetent?”
I’m sure they had a reason too. It was cheap and it seemed like the thing to do at the time. They probably thought that x86 could handle more load than they would ever have. After all, they were entering a market that was already dominated by Yahoo and other search engines. Who would have thought Google would become the most popular search engine in the world given the market they were trying to penetrate? They seriously underestimated how big Google was going to get and that in turn caused them to seriously underestimate the computer horsepower that was going to be required to run it.
I’m not saying they are stupid. I’m just saying they made a serious estimate error in how popular their service would become — probably an error almost anyone would have made given the market they were trying to penetrate. They probably thought “it’s cheap. And it will handle the load”. I don’t think anyone, including Google’s founders, had any idea how big Google would become.
I’m not using Google as an example of incompetance. I’m just using it as an example to show that x86 hardware doesn’t scale well.
Hey, pal, when you spew forth unsubstantiated nonsense like that what do you expect me to say?
Did you buy the book yet?
Google and many other companies fall into the same trap on software and hardware. They see an initial price tag and forget to factor in annual support and replacement costs. In the case of google, they started out when the whole web was relatively young, from a commercial perspective. A lot of companies don’t factor in the cost of their IS department when it comes to maintenance of these “low cost” solutions. Their rationale is that they were paying for those people anyway. This of course, ignores the fact that they probably need more of those guys, and may even have to pay them overtime to keep their system up and running.
The bottom line is that it isn’t done out of incompetence, it is done out of short-sightedness.
Damn, Hank. That’s an excelent explanation.
Can I use it
>what do you expect me to say?
I expect you to keep your thoughts to yourself if they cannot be useful like in this case
Sure Ruprecht. I’d love to get a case study to back it up too. I’m sure we can find one somewhere. That is not to say that Linux clusters *never* have a place in industry. They most certainly are very acceptable in many cases. However, they are definately *not* a one size fits all solution.
I’m NOT saying that linux/x86 clusters do not have a place. It has been proven that are quite good for tasks that can be divided into discrete processes (render farms, for example). My problem is, when people say “x86 is JUST AS GOOD AND CHEAPER than IBM/SGI/Sun”. That is such an ignorant statement. These people are obviously so incredibly un-educated….anywho…I dont want to go into a rant
That post just hit me like something that belongs on ./
Keep my thoughts to my self?! It wasn’t an *opinion* I was expressing. It was a FACT. The FACT that you know NOTHING about anything outside of the x86 universe. Before you go spouting off about ANYTHING related to hardware outside of Intell/AMD get a fucking clue about the rest of the computing industry. And as for not being helpfull? I INCLUDED A LINK ON A BOOK ON DESIGNING CLUSTERS FOR FUCKS SAKE!!! How much more helpfull do you want me to be?! Do you want me to come over to your house and beat the knowlege into your head with a fucking lead pipe!?
someone didn’t get laid