Home > Windows > Microsoft delays supercomputer version of Windows Microsoft delays supercomputer version of Windows Eugenia Loli 2005-04-06 Windows 37 Comments Microsoft said late Tuesday that it has pushed back the expected launch of a version of Windows designed for high-performance computer clusters. About The Author Eugenia Loli Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker. Follow me on Twitter @EugeniaLoli 37 Comments 2005-04-06 5:29 am Anonymous i have a strong feeling that when this comes out everyone in the IT departments will just start laughing to the point of death from how much it just sucks compared to the power of a unix variant 2005-04-06 5:35 am Anonymous it ever comes out 2005-04-06 5:35 am Anonymous It’s not like it will be a regular NT kernel. I would suspect that it would be very competitive. I think it will be widely adopted because windows are’t bad systems–it’s who runs maintains them that make them bad 😉 2005-04-06 5:35 am Anonymous i have a strong feeling that when this comes out everyone in the IT departments will just start laughing to the point of death from how much it just sucks compared to the power of a unix variant Well, while we’re predicting: The unix-knowledgeable won’t try it since it is a waste of time…those who drank the Microsoft Koolaid will brag about it (and not try it either)…and a few poor sobs will be sucked in to using it by well meaning but inept technical managers. 5 years from now Microsoft will have a product actually worth using and it will have a feature set surprisingly like most of the Unix-variety systems. Microsoft will brag about being in the top 5 though the effort and expense Microsoft expends to get there won’t make any practical sense. This bragging will sucker smaller more easily influenced companies into buying MS clustering systems. 2005-04-06 6:01 am Anonymous unfortunately, you’re probably right. there’s no other company that can continually to fail and buy its way out until enough people are suckered in to sustain the product. Windows CE/PocketPC, Xbox, IIS, MSN Search, Office, and Internet Explorer were all slow starters, but MS has enough money to keep them going when most companies would have to abandon them. eventually they do something good enough to elicit something to make it worth doing and they virtually stop development and sit back and rake in the money, change the name/version number, and rake in more money. Office 6, Office ’97, 2000, XP, 2003, virtually the same product, IE 4, 5, 6 (and sounds like 7) virtually the same. Hey, if you want a cluster that’s just going to stagnate for years until its financially beneficial for someone else to update it. by all means, get suckered in. i can’t imagine why research organizations would want to choose a cluster OS they can’t tweak. 2005-04-06 6:17 am Anonymous I fail to see how a cluster OS can still be called “Windows”. I know, it’s their flagship OS trademark, but think about it: Windows Server 2003 can keep the “Windows” name because it is a GUI-driven system; however, there is no need for a GUI on the nodes of a cluster, and I’m sure they aren’t going to just tweak the Server 2003 kernel a bit more and leave the GUI in. Microsoft is a big company with many, many creative employees; I’m sure they can come up with a catchy name. For example, Hotmail. Except for the really prudish, that is a cool name for an email service. I’m hoping there will be a similar brainstorm for this project, and put to rest any ludicrous claims that they will essentially be running WinXP on a cluster. 2005-04-06 6:20 am Anonymous I don’t know, but I have a feeling that they are on to something. The slap in the face was when Apple anounced thier own super-computer. Microsoft is alot of things, but afraid of a fight , isn’t one of them. I think that they will come up with a very user friendly os, and one that would be fairly easy to configure. Whether or not it will be extremely secure or not, I’d have to say they are making real strides there too, as 2003 was quite a bit more secure , right out of the box. And all this from a non-fanboy, who slammed them to whomever would listen. It is also not surprising that it’s beeing pushed bacl , either. It’s a huge undertaking to moove that much data at once, let alone if you haven’t ever done it before. I still believe that it will be fairly good when it does come out. What realy interest’s me is going to be a comparison of cost. If they can do it cheaper total over all cost, then they will have indeed slap the face of Apple. They have managed quite successfull in doing this in the past, so it wouldn’t shock me. As Unix, Apple, and Linux have already proven themself in all areas of super-computing, it’s time for Microsoft to step up to the plate. After all, don’t we Unix fans always say that a little competition never hurt anyone? It’s going to get real interesting around here! 2005-04-06 6:20 am Anonymous Microsoft having trouble scalling to other platforms? It ain’t a one desktop world any more kids. 2005-04-06 6:24 am Anonymous From the article: Microsoft has not announced how much it will charge for the Compute Cluster edition, but did say last month that additional computers, or nodes, of a cluster will be priced at some discount. The initial version is designed to replicate many of the features that would be offered if someone were clustering machines using Linux. Okay, so they are going to “replicate” many, but not all, of the features of a Linux cluster. That’s fine, copy away. But they are going to charge money for most of the features of a free OS? How in the world do they expect to actually sell this? Discount or not, you can’t beat better functionality for free! 2005-04-06 7:12 am Anonymous the system can be gui driven because there has to be a master node for administrators.. duh. 2005-04-06 7:37 am Anonymous As far as academic research goes (unis) this won’t fly – unix variants do it just fine and “authorities in the field” love them (or at least know them). Industry is another matter altogether. We all know how susceptible suit-types are to mesmerizing charts and presentations showing them huge savings in one of those abstract corporate ways! What I fear the most is the combination of two, like some inc. and MS strategically ganging up to “donate a classroom” to some college. We had one of those 10 yrs. ago or so and I am very grateful to it – it made me discover a hidden SGI Indy (IRIX) in the basement on which I worked from then on – now *that* was a machine! 2005-04-06 7:43 am Anonymous By anyone who has a slight clue about how MS “launches” products… so basically, everything is just fine + in schedule… 😉 2005-04-06 8:04 am Anonymous It has to be open source, people who invest in that expensive equipment aren’t going to run a stock kernel on it, they’re going to make as much fine tuning as possible. 2005-04-06 8:26 am Anonymous Seeing as how Apple doesn’t charge a per-seat license from the users of its XServe, I’m intrigued how Microsoft is going to beat that one. Keep in mind that although PCs in general have always been cheaper than Macs [you can now convert a weeties box with the innards of a Zippo lighter to start your own home-grown blade server for a tenth of the price of half a pack of gum], it wasn’t actually Windows that was cheap, was it? They don’t call it the Microsoft tax for nothing. And did I not hear people complaining that the biggest cost of a PC was the Windows license? So Microsoft has to fight against that legacy and spend a few years cranking their product up to a level where it doesn’t offend the serious administrator for even being there. Ballmer said of Windows 3.11 that it was probably one of their worst products they ever made [just paraphrasing you, Steve]. And that was after they had hyped the concept so much that people started believing Microsoft had invented the concept [and no, I am so not diving into the Xerox argument]. So, there will be the early adopters who will go through their own private purgatory while the people using the real stuff are going to ask themselves why they even bother. “Microsoft Cluster Fast Upgradable Computing Kernel has encountered an error and needs to shut down. Microsoft would like to learn more about this problem. Would you like to send a log file ? [Yes] [No]” 2005-04-06 8:38 am Anonymous I’m sure they can come up with a catchy name. For example, Hotmail. MS bought Hotmail, they did not create and name it. 2005-04-06 10:01 am Anonymous windows are’t bad systems–it’s who runs maintains them that make them bad 😉 That’s exactly my probleme with MS: If the create bloated software and overly complicated concepts which looks nice on paper but are not practical, people think they are just too stupid to handle the software. Go into a library an have a look at the bookshelf in the Computer department. Enormous amounts of read-once-throw-away-when-the-next-version-comes literature about MS products. Is this a sign of quality? How would you label your tv, stereo, car, toaster, etc., if you had to do intense research about the inner workings and constantly patch the things just to keep them going? Good? It’s the people deciding to buy crap based on marketing promises that makes IT bad.. 2005-04-06 12:21 pm Anonymous It has to be open source, people who invest in that expensive equipment aren’t going to run a stock kernel on it, they’re going to make as much fine tuning as possible. The dominant OSes on equipment that expensive are all closed-source. Linux etc. aren’t as big of players in the supercomputing world as we’d like to think. (Linux running as a guest OS under VM on an IBM Z-series does not count.) 2005-04-06 12:23 pm Anonymous Okay, so they are going to “replicate” many, but not all, of the features of a Linux cluster. That’s fine, copy away. But they are going to charge money for most of the features of a free OS? How in the world do they expect to actually sell this? Discount or not, you can’t beat better functionality for free! Because it is Microsoft. It’s sad but there are a lot of people that live and breath MS and wouldn’t consider anything else. They consider it the “safe” choice. Plus, don’t write MS off because they don’t have the best technical product. When they decide to go after a market they usually get it. Now my opinion – “I hope they fail” 2005-04-06 12:24 pm Anonymous Yeah whatever, I just wanna stick it on my home pc. I remember when I put Windows Advanced Server on my desktop – haha, the OS laughed at my pc specs. Then I tried to install it on my laptop… Whatever happened to Windows DataCenter Edition? I thought that was supposed to be their top dog OS. 2005-04-06 12:35 pm Anonymous Instead of the BSOD, it will leave the BCOD (Blue Crater of Death) 2005-04-06 1:21 pm Anonymous They’ll be in a market where they’re trying to convince people to switch which involves porting code. Apple is, but Apple is a lot easier to port to! Maybe they’re going to try and be uber-compatible? 2005-04-06 2:51 pm Anonymous I think its foolish to discount Microsoft in this arena. Microsoft has a knack of kicking everyones butt in some things. Besides they have some of the smartest developers in the world. 2005-04-06 4:50 pm Anonymous MS will learn from this how to provide high uptime on a single computer, because the only allowed reason for a cluster node to fail is hardware failure. MS will learn, and their stabilisation process will make their desktoop and server products to have higher uptime too. It is like it always was, MS would still produce uptime like Windows95 if FOSS had not stepped on the stage that massively. I think lots of people now using IIS on some Windows Server will scale up their web services with the Windows for cluster thingy 2005-04-06 5:34 pm Anonymous >How would you label your tv, stereo, car, toaster, etc., if >you had to do intense research about the inner workings and >constantly patch the things just to keep them going? Good? “Linux” 2005-04-06 6:22 pm Anonymous The dominant OSes on equipment that expensive are all closed-source. Linux etc. aren’t as big of players in the supercomputing world as we’d like to think. (Linux running as a guest OS under VM on an IBM Z-series does not count.) An IBM zSeries isn’t a supercomputer, it’s a mainframe. They are completely different varieties of computers. Your confusing them basically completely illiminates any credibility the rest of your comment may have had. But, if you’d like to try again, go over to the Top 500 list of supercomputers, find one that is said to be running Linux and then let us know why that isn’t true. Otherwise, I’m going to just go with what the Top 500 says, and that says that Linux is running well over 100 of them, which would make it a pretty big player in the supercomputer world. 2005-04-06 6:41 pm Anonymous Windows XXL 2005-04-06 7:02 pm Anonymous MS bought Hotmail, they did not create and name it They have bought it, but when they did so, it was called freemail. Later they renamed it to hotmail. 2005-04-06 8:18 pm Anonymous Microsoft bought Hotmail in 1998. It was called Hotmail before and after Microsoft bought it. http://news.com.com/Microsoft+buys+Hotmail/2100-1033_3-206717.html 2005-04-06 8:39 pm Anonymous <sarcasm>Oh … um … *gasp*</sarcasm> 2005-04-07 12:08 am Anonymous No they don’t, they lost most of the really good ones years ago. With things like cutting benefits to increase profit they won’t keep the few they have left. http://dotavery.com/blog/archive/2004/07/07/1568.aspx I’ve known a few that have worked for them, and they pretty much said they would work at McDonalds before going back. 2005-04-07 12:45 am Anonymous I used hotmail in 1997. Its name IS hotmail then and now. Why you huy insisted it is M$ who gave this name? Did you bet 10000000000000$ for this with somebody? Or you just eat too much fat? Sorry for you, M$ brain-washed. When you repeat worshipping M$, please show your proof first. 2005-04-07 2:04 am Anonymous Why would anyone serious want to use Microsoft Windows on a high-performance cluster? Any type of Linux e.g. the standard Beowulf setup is faster, more flexible, more secure, and virus-free. 2005-04-07 6:34 am Anonymous MS will learn from this how to provide high uptime on a single computer Yeah. Sure. They already learned from Windows CE how to make their OS smaller and less bloated. AFAIK, their NT kernel was derived from VMS and obviously they haven’t learned much from it… 2005-04-07 10:19 am Anonymous Microsoft should open their Souce code to Client using SuperComputers for scientific research. This will help both Microsoft and Clients to fine tune the OS for futher improvements. This is how Microsoft can actually compete with Unix boxes, otherwise it is difficult to make a dent in this highly specilized marketplace. 2005-04-07 2:28 pm Anonymous To everyone who think that Windows is inferior to Linux/Unix in the cluster space: I recently learned that the world’s largest data processing company in its industry (which also happens to be one of the worlds largest industries in terms of revenue) is also the worlds largest user of Windows based clusters. In fact, 95% of their nodes run Windows XP. And why? Because they have, over the years, found that Windows XP offers *better performance* than Linux. My company is a developer and distributor of cluster based software, so I know very well what I’m talking about here. We’re by and large a Linux boutique, however we look forward to seeing a Windows OS for clusters. For a huge number of clients (small-medium size companies), that would be the OS of choice for clusters. Yes, it would have to have a high performance MPI stack, a user friendly and powerful management and scheduling tool, but considering the growth in use of COTS clusters, I don’t see why that couldn’t happen (also) on the Windows OS. 2005-04-07 2:36 pm Anonymous To anonymous above: >Why would anyone serious want to use Microsoft Windows on a high-performance cluster? Any type of Linux e.g. the standard Beowulf setup is faster, more flexible, more secure, and virus-free. I don’t know what area you are working in, yet as often the case is : Companies doesn’t want MORE FLEXIBILITY, just MORE SIMPLICITY. Beowulf (or any other Linux OS flavour) setup isn’t very fast, it normally takes hours on a medium large (e.g. 256node) cluster. Flexibility you don’t really want – you just want it to work out of the box. Security? Who the hell cares about security on a cluster? They are all in secure rooms, the only connection to the cluster is through the Master node (or job scheduling system for end users), so you don’t need security. Virus? Most of the communication between end clients and a cluster is based on proprietary TCP protocols, so little chance of viruses getting in there.. To sum up : Businesses want more choice in terms of OS, less choice in terms of the underlying OS flexibility. Just plain simplicity, stability and performance which also implies quick turnaround time for starting new jobs, updating the OS (or even changing the OS). CEO 2005-04-09 6:39 am Anonymous Ok, a few things. First of all, even if M$ does put out a decent clustering system and sells it cheap dosen’t this remind any of you of the past….. M$ used to be cheap,…. they cornered the market, got people locked in, than jacked up their prices. Next, there is a reason that so many clustering systems and supercomputers run an opensource os, it is primarely because they can edit the code to make it do what they need, not price. (does anyone think M$ will release its sourcecode?….. HAH, … that’ll be a cold day in hell) I also may add… I am not anti Microsoft persay, although, I do think they put out a low quality product that thrives and possably survives on having people locked into it. People don’t use Microsoft because it is great, they use it because of industry standards that lock them into an inferior product.