Posted on 05/14/2008 6:55:59 AM PDT by ShadowAce
Isn’t there some sort of nerd board here at FR that this can go on rather than waste space here with it?
You don’t think NYSE news is worthy of a business/economy ping?
IB4GE
Good for Red Hat, but I’m still not sure I’d trust something like this to anything but big iron.
its a tech ping....
...whatever.
I dunno...sounds like they’re clustering, and Google manages OK. I think there are advantages and disadvantages to both approaches.
Crazy huh? Those communists at the NYSE! So can Castro now claim Linux is part of the capitalist imperialist conspiracy because it’s being used to run the biggest capitalist market in the world? It’s the same logic as saying it’s communist because Castro and Chavez prefer it.
Besides, what's the concern about space? This isn't the second millennium B.C., when people wrote on potshards because paper was a premium item. FR is not written on papyrus, but hard drives, which get bigger all the time. Just the other day I saw a prediction that within 5 years, iirc, data storage manufacturers will ship a yottabyte of capacity in one year.
So there ought to be room enough.
It will be interesting to see what happens the first time they have a failure that causes an outage.
Google doesn't actually care much about losing data or incorrect calculations due to freak processor error. Mainframes actually run each instruction twice and compare the results to ensure everything calculated perfectly. You use them when there can be no error, and I think that applies to the NYSE. But then they did this move, and they probably considered that and decided the risk was acceptable.
You may be onto something there. I peeked inside my computer and there was nary a scribe in sight. Who knew?
Trillions of dollars aren't at stake here. And Linux does run on mainframes.
You’re right. FR bandwidth is such a precious thing that we ought to use what little we have available most wisely.
>>Google doesn’t actually care much about losing data or incorrect calculations due to freak processor error. Mainframes actually run each instruction twice and compare the results to ensure everything calculated perfectly. You use them when there can be no error, and I think that applies to the NYSE. But then they did this move, and they probably considered that and decided the risk was acceptable.
Financial exchanges are more IO-bound than CPU-bound; they just don’t do a lot of fancy calculations. It’s more important for them to cluster for parallelism and risk mitigation so they don’t have a single point of failure.
Preach! ;-)
This is exactly what mainframes are good for. They have massive I/O for one. But even if the calculations aren't fancy, a processor error will make the answer wrong, and that won't happen on a mainframe. It retries the instruction, and if the error still shows it locks out that processor module and notifies that it must be replaced.
It’s more important for them to cluster for parallelism and risk mitigation so they don’t have a single point of failure.
Mainframes are internally parallel, with hot-swappable processor and I/O modules. In addition to that, you can run two IBM mainframes dozens of miles apart as failover or in parallel, and that also makes for a great disaster recovery plan (they're still running perfectly if the NYSE gets blown up).
Also, if you need to run multiple OS images (a feature of clusters of course) then use LPARs, which separate the different running OS images so well that they are considered as separate physical machines from a security standpoint.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.