×

Welcome to the Slashdot Beta site -- learn more here. Use the link in the footer or click here to return to the Classic version of Slashdot.

Thank you!

Before you choose to head back to the Classic look of the site, we'd appreciate it if you share your thoughts on the Beta; your feedback is what drives our ongoing development.

Beta is different and we value you taking the time to try it out. Please take a look at the changes we've made in Beta and  learn more about it. Thanks for reading, and for making the site better!

SGI Acquires Linux Networx Assets, LNXI Dead?

kdawson posted more than 6 years ago | from the tags-involving-cluster-not-welcome dept.

Supercomputing 96

anzha writes "It seems that that Linux Networx, the pioneering Linux supercomputing company, has gone belly up. SGI announced that it has bought the core assets of LNXI. Furthermore, the rumors are that the doors were locked and employees were just given their paychecks. This analysis, on the other hand, claims that SGI has 'made employment offers to many LNXI engineers.' It's unclear what kind of support will be extended to customers of LNXI's Clusterworx Advanced products. What does this mean for the future of Linux supercomputing?"

cancel ×
This is a preview of your comment

No Comment Title Entered

Anonymous Coward 1 minute ago

No Comment Entered

96 comments

"super"computing (2, Insightful)

Gothmolly (148874) | more than 6 years ago | (#22433782)

Isn't grid supercomputing best done on Macs, or via MOSIX, or a homebrew Beowulf cluster (insert joke here). I don't see this as enough of a commodity product to support a reseller-type market. If you need a linux supercomputer of OTS gear, you probably have enough eggheads on staff to build and support it.

Re:"super"computing (1)

Mr. Jaggers (167308) | more than 6 years ago | (#22437608)

Might be tough for the resident eggheads to spec, build & support a cluster of over 1100 nodes. How many MSCEs and MSCITPs know how to properly rack up & cool that much hardware?

Re:"super"computing (1)

spun (1352) | more than 6 years ago | (#22438292)

I worked for a company that built supercomputer clusters for Los Alamos & Sandia National Labs. Those labs have more than enough eggheads to do it, but they are real high level eggheads. They have more important things to do. Building & supporting compute clusters is one step above monkey work, why should they waste their time on it?

Re:"super"computing (1)

Alpha830RulZ (939527) | more than 6 years ago | (#22439654)

It's also not that hard to do in this day and age. We're doing one now. It took us a few weeks to set up the gear, implement grid gain (www.gridgain.com), and set up a proof of concept architecture. Now all we have to do as add nodes as required. I did it all myself. I'm competent, but not magical.

Belly Up? (3, Insightful)

eldavojohn (898314) | more than 6 years ago | (#22433810)

"It seems that that Linux Networx, the pioneering Linux supercomputing company, has gone belly up."
What causes you to think that? Have they filed for bankruptcy? Is there some indication they were failing?

According to most definitions of 'belly up' [wiktionary.org]:

1. (idiomatic) Dead or defunct, often used with go, went, or turn. (see go belly-up)
After several financial failures, the organization went belly up.
I'm pretty sure that since SGI has slowly become a niche provider for creating solutions for a few specific customers, they see Linux Networx as another good partner in another niche market. SGI isn't at the greatness they once were but it looks like they're holding their own in what they are doing.

Re:Belly Up? (1)

Vectronic (1221470) | more than 6 years ago | (#22433846)

Yeah but thats not very exciting is it?... "Just another merger"... but the more you can make it sound like the entire future of Linux depends on this single deal, the more attention you get...

Re:Belly Up? (1)

Doc Hopper (59070) | more than 6 years ago | (#22435678)

I just passed their headquarters this morning. There are no cars in the parking lot on a regular workday, and I don't recall any there on Wednesday, either. Perhaps that is an indication of their current position?

I interviewed for a sysadmin job with them several years ago and had done some consulting work on Bugzilla for them in the past, and they seemed like a good place to work. They had plenty of customers and pretty solid technology for managing clusters. But once you've deployed a few dozen supercomputers based on commodity components, and all your customers realize the only value you offer is the very-expensive cluster management software, what's your next plan to make money once they decide to save that money and buy more hardware instead?

Re:Belly Up? (1)

sl3xd (111641) | more than 6 years ago | (#22439750)

Depends entirely on the time of day you drove past. Typical working hours for most employees was shifted a few hours later in the day. There were plenty of vehicles there yesterday, (and around back), as well as quite a few today.

One reason few cars were there is they were trying to plow the parking lot in the morning (yesterday and today). There was a pretty wicked snowstorm the night before, making the commute painful.

That and most of the staff rolls in later in the morning, and leave later at night than most office drones typically do.

Re:Belly Up? (1)

bbice (1246198) | more than 6 years ago | (#22554044)

It also depends on what site/office you're talking about. The parking lot of the Sunnyvale office fills up fairly early. (shrug)

Re:Belly Up? (1)

Hognoxious (631665) | more than 6 years ago | (#22437214)

but the more you can make it sound like the entire future of Linux depends on this single deal, the more attention you get...
You must be new here...

Re:Belly Up? (2, Insightful)

neurovish (315867) | more than 6 years ago | (#22433872)

"It seems that that Linux Networx, the pioneering Linux supercomputing company, has gone belly up."
What causes you to think that? Have they filed for bankruptcy? Is there some indication they were failing?

According to most definitions of 'belly up' [wiktionary.org]:

1. (idiomatic) Dead or defunct, often used with go, went, or turn. (see go belly-up)

                    After several financial failures, the organization went belly up.
I'm pretty sure that since SGI has slowly become a niche provider for creating solutions for a few specific customers, they see Linux Networx as another good partner in another niche market. SGI isn't at the greatness they once were but it looks like they're holding their own in what they are doing.
Since SGI hasn't turned a profit in forever and usually loses about $100M a year, I'd say that having your assets bought by them would qualify you for dead.

Re:Belly Up? (5, Interesting)

eldavojohn (898314) | more than 6 years ago | (#22433920)

Since SGI hasn't turned a profit in forever and usually loses about $100M a year ...
False. Their net income [google.com] for 2006 was -$146.19 Million while their net income for 2007 was $222.61 Million. You may have been correct but at least in 2007 it looked like they have turned things around.

Re:Belly Up? (1)

neurovish (315867) | more than 6 years ago | (#22434128)

Since SGI hasn't turned a profit in forever and usually loses about $100M a year ...
False. Their net income [google.com] for 2006 was -$146.19 Million while their net income for 2007 was $222.61 Million. You may have been correct but at least in 2007 it looked like they have turned things around.
Weird, the statement I'm looking at shows a -$103.64 million income for 2007 http://finance.yahoo.com/q/is?s=SGIC&annual [yahoo.com], but their debt also substantially reduced that year. It looks like google figured that into their income calculation.

Re:Belly Up? (2, Informative)

Sen.NullProcPntr (855073) | more than 6 years ago | (#22435040)

False. Their net income [google.com] for 2006 was -$146.19 Million while their net income for 2007 was $222.61 Million. You may have been correct but at least in 2007 it looked like they have turned things around.
Not sure where google gets their numbers from but you shouldn't believe everything you read on the interwebs;-) If you go to the source [sgi.com] you will see nothing but net losses for FY2007 [PDF] [sgi.com] (which ended on June 30, 2007) and FY2008 [PDF] [sgi.com] (which somehow ended December 29, 2007!!!).
The change in FY dates may have caused the confusion in the totals.

Disclaimer: I buy high and sell low.

Re:Belly Up? (0)

Anonymous Coward | more than 6 years ago | (#22435292)

Oops -- It was SGI's Q2 2008 that ended 29DEC2007 not their FY2008. But the point is still valid - losses in each of the past 4+ quarters.

Re:Belly Up? (1)

flaming-opus (8186) | more than 6 years ago | (#22435378)

They made money on their cash ballance, and by offering more shares, that's not really the same as an operating profit. Investors aren't going to keep dumping money into you, if you don't make a return on that money better than a t-bill. (that said, they sure do keep throwing good money after bad at sgi.)

2007 may have been better than 2006 for SGI, but I look at it as loosing money on every deal, just to drive up volume. It's too bad, as they have good technology in the altix and cxfs, but they just can't seem to produce them at a price that allows them to be profitable.

Re:Belly Up? (0, Flamebait)

mjmartin_uk (776702) | more than 6 years ago | (#22435404)

Wow, SGI actually managed to scrape up some cash to buy someone! Didn't see that coming at all.

Re:Belly Up? (1)

fm6 (162816) | more than 6 years ago | (#22440510)

You're out of date. SGI is no longer big enough to lose $100M a year. They scaled way back before emerging from bankruptcy. They're now a small (but modestly profitable) company.

Re:Belly Up? (1)

jrumney (197329) | more than 6 years ago | (#22433890)

"It seems that that Linux Networx, the pioneering Linux supercomputing company, has gone belly up."
What causes you to think that? Have they filed for bankruptcy? Is there some indication they were failing?

The fact that the employees turned up to find the doors locked and received their final paychecks suggests that the company was wound down and the assets sold, rather than being sold as a going concern.

Re:Belly Up? (1)

mihalis (28146) | more than 6 years ago | (#22433964)

"It seems that that Linux Networx, the pioneering Linux supercomputing company, has gone belly up."
What causes you to think that? Have they filed for bankruptcy? Is there some indication they were failing?
The press release says SGI acquired the core assets of Linux Networx. That phrase normally refers to buying parts of a defunct company. It does not mention taking on LNXI's staff

Re:Belly Up? (1)

TheRealMindChild (743925) | more than 6 years ago | (#22434186)

... And it isn't because SGI has nothing to offer (unlike SCO). It is just what they have always done has practically become irrelevant. They made powerhouse workstations with special top-grade hardware, and software optimized for said hardware. The problem, the same problem that Sun has needed/needs to tackle is that x86 PCs are so dirt cheap, that even if they aren't the best of the best, you can just throw more hardware at it for a mere pittance.

Re:Belly Up? (0)

Anonymous Coward | more than 6 years ago | (#22434658)

SGI has been out of the workstation business for years. SGI makes large, cache-coherent non-uniform memory access systems. The do very well on large, non-parallel computing problems where a cluster just won't do.

Re:Belly Up? (0)

Anonymous Coward | more than 6 years ago | (#22434936)

well as a past employee of LNXI, Bo did put on a good show, and gave his friends jobs there, and made it a nice place to sit at the top. Its great to be connected right? So after spending lots of our money, not supporting development and just plain making it a place where.....it was a festival of bad decision making.....gets rid of the people that built it to what it was, and then leaves as well. Isnt it great about how lousy management seems to never matter, you can always get another high paying job and then attempt to destroy the company...or sell it off. LinuxNetworx was once a fantastic place full of innovation and energy. Sure it was small, quaint and fun even back to the days at the dogpound. It was a family. But when the money started to flow in from the sharks, it was the beginning of the end. Farewell old friend...farewell.

Re:Belly Up? (1)

Falcula (20022) | more than 6 years ago | (#22434956)

What causes you to think that? Have they filed for bankruptcy? Is there some indication they were failing?
Well, the first bad sign I noticed is when they were a no-show at the SuperComputing 07 conference last November. They had a booth, and were noted in the program, but had pulled out late.

That and their website hadn't changed in months.

Teh Lunix can never fail (0)

Anonymous Coward | more than 6 years ago | (#22435188)

Teh Lunix can never fail, it can only BE failed.

These guys obviously weren't doing it "right", because everyone knows that spending all your time working on something you give away for free is the path to wealth and fame.

The future of Linux supercomputing (5, Insightful)

jrumney (197329) | more than 6 years ago | (#22433830)

What does this mean for the future of Linux supercomputing?

It means the future of Linux supercomputing will be backed by SGI. You don't think SGI bought an already dead company just to kill it, do you?

Re:The future of Linux supercomputing (2, Interesting)

Maller (21311) | more than 6 years ago | (#22435626)

One Linux-based HPC vendor bought out another Linux-based HPC vendor, this will not really effect Linux on HPC. All of the Top 10 of the TOP500 use Linux in one way or another. The Blue Genes have SLES on the service nodes and CNK on the compute nodes. The SGI is SLES with add-ons. The HPs are Linux clusters. The Crays are SLES on the service nodes and either Catamount or Linux on the computes.

Linux is very pervasive in HPC and becoming more so. Since I know a little something about Cray, the newest vector and scalar lines both use Cray Linux Environment (CLE) formerly called Compute Node Linux (CNL). Cray's CNL was released in second half of 2007 and already over half of the Cray processors in production are running CLE on the computes instead of Catamount, the very lean, proprietary compute node OS.

Re:The future of Linux supercomputing (0)

Anonymous Coward | more than 6 years ago | (#22436150)

All the small Linux HPC commodity cluster companies are having a very hard time... none make a profit, all are VC capitalized, some on double-digit rounds of funding (LNXI was on its ~4th round, I believe). The original CEO of LNXI (the Company was formed around Transputers in 1989 as Alta Technology), Glen L. said that "one day the Linux Cluster market will be owned by the large companies, Dell, IBM, etc...". It looks like he was right, as that's where the market is now, and these companies can validate any losses in HPC w/ their overall strategies.

SGI itself is a startup from 2006 (when they nullified all their existing stock and started from scratch funding) that acquired the name of the former supercomputing giant and some skills in big-iron HPC... but haven't been able to sell that profitably.

They've begun selling commodity clusters, but their big-iron bend doesn't allow them to understand commodity margins, and Bo E. (SGI's CEO, who helped kill LNXI too, as a former CEO of LNXI) still doesn't understand commodity-based clusters.

I have little hope for the company that calls itself SGI.

Good luck to all the folks that got their final paychecks yesterday... it was a valiant effort on your part tied down by too many years of crappy management.

Re:The future of Linux supercomputing (1)

jd (1658) | more than 6 years ago | (#22437622)

OpenMOSIX revival attempts were badly derailed by folks who saw it as only a load-balancing solution that could (and should) never be used for HPC. I took - and take - the line that a good design solves many problems and have considered whether forking OpenMOSIX or developing an alternative would be the better way to go. Linux HPC has many, many tools but very little imagination, and that has poisoned it to a degree.

You can use one of many variants of scalable reliable multicast to deliver data to many places in a single transaction, you can use RDMA to directly transfer to/from memory without CPU intervention, you can use scatter/gather or virtual memory management to handle discontiguous data. But you can't do any two of those at the same time. Logically, these are wholly independent and shouldn't need to know or care what else is done, but semantically, that's not how they're written.

There are cluster-wide schedulers, grid schedulers, meta-schedulers, all open-source and all designed to allow a cluster or grid to unify activity, optimize the use of resources, take advantage of the fact that with more knowledge comes better utilization of time. How many Linux clusters make use of such tools?

This isn't a slam on Linux. It's not the fault of Linux that nobody uses utilities that exist for it. Particularly as nobody seems to use those utilities anywhere else, either. HPC potentials lie decades ahead of HPC realities and the gap is getting wider, not narrower. Not because people are developing possibilities so fast, but because the realities are changing so slowly.

Re:The future of Linux supercomputing (0)

Anonymous Coward | more than 6 years ago | (#22444846)

Logically, these are wholly independent and shouldn't need to know or care what else is done, but semantically, that's not how they're written.

Logically, you are correct in your argument, and props for that. But watch out, you are muddling/poisoning the meaning of "semantics". Look at what it covers in linguistics (where it comes from) and try opposing "logic" and "meaning"... And if you expand from linguistics to semiotics, you'll often see logic wholesale equated as semiotics, the study of all signification and communication, and you are welcome to try dealing with that without meanings all over the place. ;-)

For your purpose here, "theoretically" (or just "as such") vs. "practically" (or "as implemented") would perhaps be a better fit.

Re:The future of Linux supercomputing (1)

jd (1658) | more than 6 years ago | (#22448260)

Ok, I agree with you. Consider strings substituted accordingly.

Re:The future of Linux supercomputing (0)

Anonymous Coward | more than 6 years ago | (#22438036)

Meh. They've done worse.

Become irrelevant? Buy out the competition! (1)

Zantetsuken (935350) | more than 6 years ago | (#22433850)

What does that mean? It means they decided they became irrelevant to the industry at some point or another, and they know they can't get back up there with superior technology and code, so they'll just buy the competition out so that people in the market have no choice but to go to them.

What? (3, Interesting)

xzvf (924443) | more than 6 years ago | (#22433870)

Weak company buys weaker company just to shut it down? Am I missing something? What percentage of the super computing market does SGI and Linux Networx have now? With the top 500 dominated by Linux systems I think Linux based super computing is in good shape. Sure customers that took a chance on Linux Networx MAY be screwed, but only because SGI isn't in a strong position to be around much longer. Someone who cares should look into the deal and the involvement of any LBO firms. Smells kind of SCOish.

Re:What? (0)

Anonymous Coward | more than 6 years ago | (#22434008)

"Sure customers that took a chance on Linux Networx MAY be screwed"

But those customers have the source code !?!, they can just hire some consultant to fix whatever problem they have, if that dont work out I'm sure their nearest LUG can help them out.

Re:What? (0)

Anonymous Coward | more than 6 years ago | (#22434458)

Someone who cares should look into the deal and the involvement of any LBO firms. Smells kind of SCOish.

Putting aside the fact that you obviously have no idea what an "LBO firm" is, you might want to wait for more than "rumors" repeated by some random blogger before panicking.

Re:What? (1)

lymond01 (314120) | more than 6 years ago | (#22435296)

SGI bought their assets which I take to mean "stuff". Servers, switches, racks, pocket protectors...

Re:What? (1)

afidel (530433) | more than 6 years ago | (#22435770)

Nah with a company like that it usually means IP including patents, trademarks and copyrights.

Dude.. wait, what? (4, Funny)

ackthpt (218170) | more than 6 years ago | (#22433926)

I thought SGI was already dead.

Re:Dude.. wait, what? (3, Funny)

Hythlodaeus (411441) | more than 6 years ago | (#22435638)

A dead company came back to life and now controls another dead company. Does that make it a lich company?

Re:Dude.. wait, what? (2, Funny)

jd (1658) | more than 6 years ago | (#22437646)

According to my Rolemaster manuals, if it's a greater Lich and it's posessed by a greater Demon (ie: switches to *BSD), it becomes a Black Reaver.

Re:Dude.. wait, what? (2, Informative)

fm6 (162816) | more than 6 years ago | (#22436444)

In a sense, they did die. They went bankrupt, and their stockholders lost their entire investment. But they had a few products worth saving (massively parallel Itanium and x64 systems), so new investors bought the name and those products. Pretty much a new company, and not a major player.

It's actually kind of similar to Cray, which SGI bought, ran into the ground, and then sold to Tera Computer. Tera did get a couple of Cray products (others stayed with SGI or had already been sold to Sun), but I suspect that Tera just wanted to rename itself Cray.

The workstations are no more, and there are no more Irix/MIPS systems. Everything runs Linux. Hence their interest in a high-performance Linux company.

Re:Dude.. wait, what? (0)

Anonymous Coward | more than 6 years ago | (#22443474)

No - you're confusing SGI with BSD

Oh fun (3, Insightful)

downix (84795) | more than 6 years ago | (#22434102)

The last time SGI bought a supercomputing company things did not go well. SGI has managed to shoot themselves in the foot constantly for over a decade. At one time, they were an industry leader (even have an Indy sitting before me now) now they're in trouble and know it. Their abandonment of MIPS and embrace of Itanium gained them short term benefits, but gutted the long term profitability and flexibility of the company. Now they're desperate for growth before the stockholders abandon them utterly.

Suggestion SGI, invest in new CPU's, the market is wide open for a solid x86 competitor now that PowerPC's given up the ghost there. Partner with Sun, use the OpenSPARC, make a consumerish-model that fits into customized Opteron motherboards, do something other than stand there admiring your own navel!

Re:Oh fun (1)

halfelven (207781) | more than 6 years ago | (#22436560)

Abandoning MIPS was what kept them alive for a while longer.
Invest in a new CPU - I don't think they can afford that now. And Sun is their traditional arch-enemy.

Re:Oh fun (1)

downix (84795) | more than 6 years ago | (#22436992)

But the main reason they got to the point that abandoning MIPS saved them is because they abandoned MIPS development upon the Intel announcement of IA64. If they had not done that, and a series of ill timed lawsuits against MIPS core vendors, they would not have gotten to that point.

Re:Oh fun (0)

Anonymous Coward | more than 6 years ago | (#22477232)

Umm, SGI already abandoned their stockholders. When they emerged from bankruptcy a few years ago the old common stock was canceled. The bondholders at that time became the new stockholders.

I cant remeber (1)

MrShaggy (683273) | more than 6 years ago | (#22434290)

Did they ever try to produce a graphics card?? It would have made sense.

Re:I cant remeber (1)

gfxguy (98788) | more than 6 years ago | (#22434372)

They did make cards for PCs before most people even considered using PCs for any sort of graphics...

The problem was the cards were nearly as expensive as buying one of their low-end workstations and didn't work nearly as well.

By the time consumer 3D cards were coming out, SGI was already on it's way down.

Re:I cant remeber (1)

MrShaggy (683273) | more than 6 years ago | (#22434724)

I knew someone that had o2 box on her desk at work. That was a long time ago. I also think that if they are looking at new markets, as they seem to be in chpt11(?). It would make sense. As long as it is competitive.

Re:I cant remeber (1)

fat_mike (71855) | more than 6 years ago | (#22441062)

We used to have a Canon CLC900 that used an O2 as the ripper. I loved that little blue toaster. It was cool looking, powerful, and it looked like a blue toaster.

I haven't had a color printer since that is truly WYSIWYG since.

Re:I cant remeber (1)

gfxguy (98788) | more than 6 years ago | (#22448160)

O2's on Ebay for as little as $30.00 now... of course, you need adapters and stuff in order to get it to work with regular monitors and so forth, but overall a pretty neat toy...

Sun to buy SGI (1)

teknopurge (199509) | more than 6 years ago | (#22434388)

Nothing I've heard, but as a Sun shareholder I would really like to see this happen. SGI is _cheap_ right now for the IP it has.

I miss IRIX too.....

Regards,

Thank god. (0)

Anonymous Coward | more than 6 years ago | (#22434422)

The fewer vendors shipping shitty Linux clusters that don't work properly the better. I'm aware this opinion won't be popular here, but Linux is the Windows of the supercomputing world - the software overpromises, underperforms and is completely unreliable.

I waste so much of my life trying to get these machines to do relatively simple things that just work on proper supercomputers that it's just not funny. Sadly universities are sucked in by these vendors who claim they can provide them with a cheap alternative to a real machine and this in turn is killing off the proper vendors while making the users' lives a living hell.

I don't suppose SGI fancies buying out ClusterVision as well, do they?

Obviously, posting anonymously because I don't want to be fired.

Their management has always been jacked (0)

Anonymous Coward | more than 6 years ago | (#22434440)

As a local, and having had connections to that company they have always had management issues. I don't know for sure, but it doesn't sound impossible to me that they closed doors.

Supercomuting is off the shelf now (3, Informative)

mlwmohawk (801821) | more than 6 years ago | (#22434478)

The economies of "COTS" "Consumer Off The Shelf" technology and the advancement of projects like MPI and PVM, as well as gigabit ethernet has made fast and effective clustering almost as easy as plugging in an Ethernet cable.

Seriously, while "programming" an application takes some chops, the infrastructure to run it is trivial.

"In my day" we had, at best, 10mbit ethernet. We had to use special drivers to get out "Dolphin Interconnects" working right. We had to really study the network topology to get the message passing right.

These days, forget about it. virtually all ethernet is interconnected via a switch so collisions are no longer an issue, switches don't cost thousands of dollars anymore, network interface cards use busmastering PCI or PCI2 (not ISA), The networks are 100x faster. The computers are 100x faster.

What's the point of a company who's products only tend to mitigate (not eliminate) the inevitable diminishing returns? Can you say buggy whip? Yea, sure, people still make them, but they are not in common use.

Re:Supercomuting is off the shelf now (1)

Wells2k (107114) | more than 6 years ago | (#22435486)

In addition, even things like Infiniband have come down significantly in price recently. It is now possible to get Infiniband HCA's (4X SDR) for $125 USD new, and an 8-port switch for $750 USD.

With that kind of pricing, it is now plausible to setup a small cluster in your own home with very high speed bandwidth between nodes. 4X SDR Infiniband is capable of sending data at 10 gigabit speeds, and running IPoIB on it gives you the same bandwidth as the much more pricey 10 gigabit ethernet that is out there.

Re:Supercomuting is off the shelf now (1)

mlwmohawk (801821) | more than 6 years ago | (#22435724)

In addition, even things like Infiniband have come down significantly in price recently. It is now possible to get Infiniband HCA's (4X SDR) for $125 USD new, and an 8-port switch for $750 USD.

I'm not saying that these things aren't useful, but the point I was trying to make is that "super computing" (well, lets call it highly parallel computing) is far better understood today than ever before and while we have these specialized high speed links, it is important to note that the trick to highly parallel "decoupled" processing is factoring in the data transit time. Bigger chunks of processing sent/received less often.

Look at Seti@home. Arguably one of the most powerful super computing platforms ever created, yet its transfer characteristics were ridiculous. But! it worked well, and the reason why it worked well was that the actual processing time far out weighted the data transfer time.

The same thing is true for the high speed interconnects. They only reduce the data transfer time between processing entities, they do not eliminate it. The speed increase in ethernet has been nothing but a boom for clustering as many of the strategies were designed around relatively slow interconnects and have thus "magically" become far better and more functional than they were with no work. This has challenged the specialized devices as their cost/benefit ratio continues to erode.

Re:Supercomuting is off the shelf now (1)

afidel (530433) | more than 6 years ago | (#22435872)

Can those $125 cards offload the IPoIB? Because trying to do TCP/IP at 10Gbit without offload is a lesson in futility. In fact even maintaining actual 1Gbit speed takes a lot of overhead unless you are using jumbo frames, which are only useful for bulk data transfers.

only for some algorithms (1)

halfelven (207781) | more than 6 years ago | (#22436626)

That works only if you can split the problem in totally independent chunks. Communication times are still very high with off the shelf components.
If the various parts of the problem require quick exchange of data very often, off the shelf computing fails miserably. In that case you need a true supercomputer, as in a single-OS-image machine, like the ones built by SGI, IBM, etc.

Re:only for some algorithms (1)

mlwmohawk (801821) | more than 6 years ago | (#22436886)

That works only if you can split the problem in totally independent chunks. Communication times are still very high with off the shelf components.

And that is the science of writing a good parallel algorithm. Not all problems can be divided into a parallel paths others require a full working knowledge of the problem.

If the various parts of the problem require quick exchange of data very often, off the shelf computing fails miserably. In that case you need a true supercomputer, as in a single-OS-image machine, like the ones built by SGI, IBM, etc.

To me, that's like using assembly language to code a brute force algorithm to obtain greater speed. I've had to do it, but it shouldn't be the default option. A finely grained parallel algorithm doesn't scale well and while using tight NUMA architectures may make a certain class of problem easier to solve, usually upon further examination one can typically develop a more distributable algorithm.

Re:only for some algorithms (1)

halfelven (207781) | more than 6 years ago | (#22437244)

You are correct, but remember: despite best efforts, it's sometimes impossible to split an algorithm into independent chunks.
It depends on the problem you're solving. Some of them are tough cookies.

Re:Supercomuting is off the shelf now (0)

Anonymous Coward | more than 6 years ago | (#22442414)

Actually, COTS stands for Commercial Off The Shelf.

FYI... (1)

Penguinisto (415985) | more than 6 years ago | (#22434538)

Linux Networx, based in Lindon, Utah (sound familiar?) was one of them thar Canopy/Tarantella -type companies (which SCO happens to be).

BUT... unlike their evil twin sister, LNXI is a pretty cool bunch of folks. I got to tour their facilities once (they were looking to contract some Linux training, and I was looking for a side job at the time. A couple of my former students ended up working there. :) ).

I gotta give 'em props... they were doing some pretty cutting-edge stuff at the time, and they probably still do. They also went out of their way to not be associated with the McBride gang, so IMHO they deserve to stick around.

/P

SGI will go bankrupt AGAIN (0)

Anonymous Coward | more than 6 years ago | (#22434600)

I have worked extensively with their gear since the early 90's and watched their stock fairly closely for the past 7 years. They recently went backrupt and their cash burn rate suggests it will happen again in the not too distant future. It was sad watching the company get driven into the ground by management but so it goes...

MINUS 1, TROLL) (-1, Troll)

Anonymous Coward | more than 6 years ago | (#22435434)

members are the resignation survive at 4al Walk up to a play

A prudent investment. (2, Interesting)

flaming-opus (8186) | more than 6 years ago | (#22435526)

SGI probably got the technology for pennies on the dollar. When a company closes its doors, the investors and creditors are left holding the bag, and they're interested in getting out from under a little bit of that debt, and do it quick. If they don't unload the intellectual property quickly, it decays, looses mindshare in the marketplace, and falls out of date. This is doubly true in the world of linux, where you have to keep up with the kernel changes, and the changing distributions.

Similarly, SGI has changed a lot of their focus from their expensive cache-coherent single-system-image servers to clusters of small/cheap nodes. SGI has great compiler technology, data-management software, and systems integration knowledge. They may not, however, have great systems-management tech. You don't need that for single-system-image machines. Even the big columbia machine at nasa is only a cluster of 20 machines. You can do a lot of stuff by hand, or with creative shell scripts, when you're dealing with 20 machines. With 400, it's tougher. I'm sure this won't solve all their problems, but I bet it will help quite a bit.

Re:A prudent investment. (0)

Anonymous Coward | more than 6 years ago | (#22437466)

Well, the weird part here is Bo Ewald's job before SGI was at ... Linux NetworX. I understand the parting
was not friendly. I also understand their recent successes came (the financials show it) AFTER he left.
The stories of Bo's reign at LNXI are ... not pretty.

So, the poor folks at LNXI, now in recovery from 'ol Bo Ewald, now find they are working for him again.

It's a sad end to an old cluster veteran. They weren't perfect, but at least they were innovative in
many ways.

Look for the combined SGI and LNXI to go into 'power dive'.

Former CEO of Linux Networx Buys Old Company (2, Interesting)

ryanisflyboy (202507) | more than 6 years ago | (#22435572)

The CEO of SGI used to be the CEO of Linux Networx:

Robert "Bo" Ewald
Chief Executive Officer

Bo Ewald joins SGI as CEO with over 25 years of relevant industry experience in the high performance computing markets. He is a seasoned industry veteran with a successful track record as a CEO.
Rather interesting, don't you think?

Re:Former CEO of Linux Networx Buys Old Company (1)

ryanisflyboy (202507) | more than 6 years ago | (#22435606)

Oops. Helps if I quote the right part:

Prior to SGI, Bo was Chairman and CEO of Linux Networx, Inc. Earlier, Bo served as President of Human Resource Solutions of Ceridian Corporation; CEO of Scale Eight, Inc, a high performance network-clustered storage company; and President and CEO of E-Stamp.

Re:Former CEO of Linux Networx Buys Old Company (0)

Anonymous Coward | more than 6 years ago | (#22435916)

And before that he was the CEO of SGI.

Re:Former CEO of Linux Networx Buys Old Company (1)

elsmob (751783) | more than 6 years ago | (#22440492)

It is no coincidence that SGI purchased lnxi. After Bo joined lnxi, he hired many current and ex-SGI people. An insider tells that many of the Bo cronies were offered jobs after the closing. It is also rumored that after Bo left lnxi, he knew the customer base of lnxi and would often try to sell to these customers.

Re:Former CEO of Linux Networx Buys Old Company (1)

ryanisflyboy (202507) | more than 6 years ago | (#22440782)

It is no coincidence that SGI purchased lnxi. After Bo joined lnxi, he hired many current and ex-SGI people. An insider tells that many of the Bo cronies were offered jobs after the closing. It is also rumored that after Bo left lnxi, he knew the customer base of lnxi and would often try to sell to these customers.
I heard the same thing. What a rotten way to run a business. Our local LUG has a small thread on the subject:
http://www.sllug.org/pipermail/sllug-members/2008-February/010039.html [sllug.org]

Re:Former CEO of Linux Networx Buys Old Company (0)

Anonymous Coward | more than 6 years ago | (#22442102)

Bo was not successful at LNXI; it was never able to recover from the damage he inflicted.

As a former SGI Employee, I'm forced to ask: (2, Funny)

cutecub (136606) | more than 6 years ago | (#22435894)

... SGI still exists? I had no idea.

Where have they been lately?

Are they cold at night?

Do they need food?

Have they been incarcerated?

Maybe I should make a donation?

... ok. Maybe I'm being a little snarky but, c'mon guys, you're keeping a really low profile for a company that's trying to sell stuff.

-S

The Cube (0)

Anonymous Coward | more than 6 years ago | (#22439136)

If you worked for SGI, did you manage to get one of those chrome-plated diecast metal paperweight / desk ornaments that was a real life 3-D model of the SGI's famous wire cube logo? I once met one of SGI's VP's back in the 1990's who had one and showed it to me. When I asked where I could buy one and how much it costs, he just laughed at me and said it was a secret. Maybe only a select few senior SGI'ers were privileged to get them, I really dunno, but I sure wanted one for myself just because it was uber-cool. You'd think that by now that at least some of them would've showed up on eBay, but I've never seen one advertised, and only ever saw that one in person.

The house that Bernard D. built and destroyed (0)

Anonymous Coward | more than 6 years ago | (#22436796)

A case study: What caused LNXI to be popular, and what caused its demise?

LNXI, as Alta Technology, was the first commercial company to sell commodity-based Linux supercomputers... that was about 1996, to Brookhaven labs. These were 200MHz PII's running RH 5.0 (the buggiest RH ever).

By 2000, the company was full of ideas of where to run with the technology, but had no R&D money: Bernard Daines stepped in w/ a capital infusion that saved the company, and allowed R&D to put its money where its mouth was, and it did a great job.

By 2002, the company had their Clusterworx provisioning/monitoring browser-based software running, along w/ their ICEBox embedded device that concentrated serial consoles over the net and controlled power and monitored temperatures independently of the motherboard (still a better ides than IPMI), and the vertical blade-like design. Simultaneously, the company was no longer producing Alpha motherboards and had abandoned this type of R&D intensive, low margin, endeavor.

This software/hardware was LNXI's mainstay through ~2005. This wave was rode to the end.

The problem was: in ~2003, Bernard took control, and in doing so brought upon yesterday's demise of the company. As poor as subsequent management was, in 2003 he crippled the company in ways that no management could have averted.

In ~2003, Bernard had decided what future R&D would take place, with two extremely poor decisions:

1) The juvenile approach to software engineering: if it works, re-write it completely from scratch, and don't learn anything from the previous version, and add in "requirements bloat" too. That was the new Clusterworx. It was to be completed in ~2003... it still wasn't finished in 2005. In the meantime, the Clusterworx that worked well was allowed to rot rather than evolve (because the new Clusterworx will be ready any day now).

2) Start making motherboards again. This R&D effort was also to be completed by ~2003, but was finally stopped in 2006 when it still didn't work.

Subsequently, there was no continuation of the previous wave, the company had no choice but to flounder.

I hear Bernard was able to get much of his money back from subsequent rounds of investors.

Good riddance... (0)

Anonymous Coward | more than 6 years ago | (#22437494)

As somebody who had to sit through a bunch of their product pitches over the last few years, I say "Don't let the door hit you in the butt on the way out."

LNXI were basically a smallish white-box vendor who wanted to charge IBM-like prices for commodity hardware strictly on the basis of their management software, which in reality was nothing special. To my knowledge, they didn't really have anybody working on storage/file systems, interconnect networks, scheduling, or any of the other really hard open problems in HPC. As a result, they were caught between mass-market vendors like Dell (who could undercut them on price any day of the week) and total-solution vendors like IBM (who did have a good story to tell about storage/file systems, interconnect networks, scheduling, etc.), with the predictable results.

How they ever convinced Boeing and several of the DOD major shared resource centers to buy their hardware, I'll never know.

We just did the last LNXI field job ever.... (1)

vil3nr0b (930195) | more than 6 years ago | (#22441892)

Well, my three coworkers and I were in Kansas City yesterday and today moving a cluster on behalf of LNXI. This was the last official LNXI job ever. Understand that 90 percent of field engineer work (cluster installations, repair, RMA work, etc.) was done by one subcontractor working for LNXI. I am not a pussy so I don't post anonymously, but I can't name it. However I was in the field acting as labor and fallout boy for the hardware and I spent a majority of my time doing Boeing RMA work for the last three years for them. I learned a lot and their technical staff (before Bo pissed everyone off) were second to none. So thanks LNXI, even though the prepaid hotel didn't work this time, we found out from the internets before you told us, and left three big node shipping crates onsite because you fired the motherfuckers who might want it back. Regardless, here comes the hate: The non technical management and venture capitalist pigs ruined a company that had a niche market creating supercomputers for the some of the best software developers in the nation. Bo ruined that by trying to compete with Dell, IBM, and the other souless fucking sellouts posing as friends of open source while forcing proprietary crap down everyone's throat. So for the next while I will assure their customers that the company that just fucked over a lot of people I admire will be fixing their stuff from now on and I am sure they will get treated better than us people. After all, I know supercomputers are worth more than staff, but for the love of God: SGI? You gotta be fscking /kidding me. For those of you who insulted our company today through snide remarks, remember you can let all the 1337 open source software and cluster companies die and noone will be left to hear you scream as IBM/DELL/HP/SUN/Microsoft/SGI drags you away.

SGI bought LNXI for the software (0)

Anonymous Coward | more than 6 years ago | (#22477350)

As the HPCwire article [hpcwire.com] noted, SGI was most likely interested in LNXI's software, particularly Clusterworx.

I know personally several Minnesota-based LNXI engineers who accepted offers at SGI and are now working out of the Eagan office.
Check for New Comments
Slashdot Account

Need an Account?

Forgot your password?

Don't worry, we never post anything without your permission.

Submission Text Formatting Tips

We support a small subset of HTML, namely these tags:

  • b
  • i
  • p
  • br
  • a
  • ol
  • ul
  • li
  • dl
  • dt
  • dd
  • em
  • strong
  • tt
  • blockquote
  • div
  • quote
  • ecode

"ecode" can be used for code snippets, for example:

<ecode>    while(1) { do_something(); } </ecode>
Sign up for Slashdot Newsletters
Create a Slashdot Account

Loading...