Weekly Edition
Return to the Press page
| |||||||||||||
Sun rises on talk of IBM deal. Good for Linux? (Tectonic)
Alastair Otter
considers
the ramifications of IBM's potential acquisition of Sun Microsystems
in a Tectonic article.
"Clearly the market likes the idea of IBM snapping up Sun but would such a deal be good for open source and Linux? Its hard to say but there are many advantages in such a deal. For a start, despite its heritage as a hardware vendor, Suns future looks certain to lie in open source software, even though it is finding it incredibly hard to make that transition. Sun owns some very valuable software properties including Java, MySQL and VirtualBox, items that IBM could well monetise if it could get its hands on them. And in doing so it might well preserve and grow these properties."
(Log in to post comments)
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 19, 2009 21:23 UTC (Thu) by jwb (guest, #15467) [Link]
Fantasy: IBM buys sun and releases ZFS under the GPL v2.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 19, 2009 21:50 UTC (Thu) by ewan (subscriber, #5533) [Link]
....and it turns out that the kernel devs won't merge something that
contains its own RAID implementation and duplicates LVM anyway.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 19, 2009 21:52 UTC (Thu) by rahulsundaram (subscriber, #21946) [Link]
Btrfs got merged
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 19, 2009 22:56 UTC (Thu) by butlerm (subscriber, #13312) [Link]
By all appearances, hybrid filesystems like zfs and btrfs are an order of
magnitude easier to set up and maintain than an LVM solution. Apparently rather more reliable too.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 13:45 UTC (Fri) by mmcgrath (guest, #44906) [Link]
> By all appearances, hybrid filesystems like zfs and btrfs are an order of magnitude easier to set up and maintain than an LVM solution.
LVM is pretty darn easy.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 6:34 UTC (Fri) by chad.netzer (✭ supporter ✭, #4257) [Link]
LVM sucks (no barriers) and block-based parity RAID suffers from the write hole. The sooner we move beyond both of them, the better off we all are.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 13:32 UTC (Fri) by michaeljt (subscriber, #39183) [Link]
Is that the only thing wrong with LFS? Sounds potentially fixable.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 16:50 UTC (Fri) by chad.netzer (✭ supporter ✭, #4257) [Link]
I'm assuming you meant LVM (not LFS)? I was a tad harsh on LVM :), because the main nice thing about it is that it is a general solution (ie. available at the block device layer, and thus available to basically all filesystems). But the truth is, you really want snapshot semantics to be at the filesystem layer, since then it can take advantage of filesystem structures. It can know what is data vs metadata, what structures are in the middle of an update (rather than just what blocks), etc. Unless you put your filesystem in a quiescent state (by remounting read-only), any block layer snapshot is as best, suspect. Same with parity raid; I don't want to rebuild my entire 100TB filesystem everytime a disk fails (or is pulled out for Seagate firmware upgrading :) ). I want to rebuild only the 1TB (ie. 1%) of data that actually exists in the filesystem. Block level RAID semantics cannot give you that, filesystem RAID can...
So block level LVM and RAID should remain, but that doesn't preclude additional filesystem implementations. Much of the code should be reusable (parity calculation, for example) so it really isn't that big a deal, imo. Start playing with btrfs snapshots, and you will be amazed at how cool they are compared to LVM snapshots.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 18:47 UTC (Fri) by michaeljt (subscriber, #39183) [Link]
Reading your reply, I realise that I completely misunderstood what LVM is (I don't use it myself). I had assumed that it was something *on top* of other filesystems, like a union filesystem, which passed on operations to the underlying filesystems and distributed files among them - perhaps this explains my typo above... That might have certain advantages for snapshotting over real LVM (I presume that snapshotting just boils down to backing up pieces of files which are overwritten), but would probably not be of any use for RAID, and would probably have lots of other downsides.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 20:58 UTC (Fri) by foom (subscriber, #14868) [Link]
> I want to rebuild only the 1TB (ie. 1%) of data that actually exists in the filesystem.
> Block level RAID semantics cannot give you that, filesystem RAID can...
Block level RAID *can* give you that, once linux and the filesystems support telling block devices
Non-suspect block layer snapshots Posted Mar 22, 2009 23:12 UTC (Sun) by xoddam (subscriber, #2322) [Link]
> Unless you put your filesystem in a quiescent state (by remounting
> read-only), any block layer snapshot is as best, suspect.
This need not be true! A filesystem with solid in-order semantics (journaled data or log-structured, most implementations of data=ordered) should make it possible to retrieve a non-suspect state from a block-layer snapshot.
The problem is analogous to crash recovery :-)
Non-suspect block layer snapshots Posted Mar 23, 2009 4:28 UTC (Mon) by chad.netzer (✭ supporter ✭, #4257) [Link]
I was really thinking in terms of data backups from snapshots; one would prefer to be in a state where not only is the filesystem safe to recover, but that all apps have flushed pending data (and are perhaps in a "recoverable" state) and are paused until after the snapshot, which represents such a "quiescent" state.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 1:04 UTC (Fri) by flewellyn (subscriber, #5047) [Link]
Or DTrace?
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 9:56 UTC (Fri) by k3ninho (subscriber, #50375) [Link]
Or Solaris' Zones (application-level virtualisation)?
I continue to be confused why the Linus-way of the middle ground between extremes isn't being taken with respect to virtualisation: full-blown virtualisation clogs up bus bandwidth with multiple full operating systems running on one machine. Lower the overheads by virtalising processes, but to do that well you'll need a good security model -- who accesses what -- and impeccable configuration tools for that security system.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 10:07 UTC (Fri) by ebiederm (subscriber, #35028) [Link]
You mean all of this container code I am running on a regular basis that has been merged into the mainstream linux kernel?
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 26, 2009 10:03 UTC (Thu) by bangert (subscriber, #28342) [Link]
Linux Containers can be found at http://lxc.sourceforge.net/
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 19, 2009 22:26 UTC (Thu) by bartman (subscriber, #3476) [Link]
There are much worse companies that could end up snatching up Sun patents.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 19, 2009 23:53 UTC (Thu) by ncm (subscriber, #165) [Link]
Good, they can finally retire Java.
LOL? Retire? I think not... Posted Mar 20, 2009 6:58 UTC (Fri) by khim (subscriber, #9252) [Link] IBM invested heavily in Java over the years so idea that IBM will kill or even put it on backburn is uttery ridiculous. If anything they want Sun for Java.
LOL? Retire? I think not... Posted Mar 20, 2009 9:57 UTC (Fri) by k3ninho (subscriber, #50375) [Link]
IBM already have an extensive licence to Java, beyond what has been GPL'd. There isn't much control over Java that IBM could gain (perhaps trade marks?).
LOL? Retire? I think not... Posted Mar 20, 2009 13:53 UTC (Fri) by jpetso (guest, #36230) [Link]
They might dual-license Java so that Apache Harmony can incorporate it.
LOL? Retire? I think not... Posted Mar 20, 2009 18:45 UTC (Fri) by man_ls (subscriber, #15091) [Link] They might want some control over the evolution of the language. Plus, gaining an insider edge on Oracle might mean billions in middleware revenue. It is indeed a bit worrying if you look at it from the POV of free software: Red Hat's JBoss might find life a bit harder too.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 12:23 UTC (Fri) by tialaramex (subscriber, #21167) [Link]
You can't retire Java now, it's maturing into a pretty decent language. I am using it for a medium-sized project at the moment, after years of managing to mostly avoid it, and somewhere in the intervening time they've fixed a lot of the deficiencies
For example, they added a printf() family and, to make that usable they added varargs. They added a natural foreach, and to make /that/ usable they implemented generics, which despite the slightly clumsy syntax finally get rid of most tiresome boilerplate casting.
These are proper language features, which didn't fit some vague hand waving vision of how programming /should be/, but just made it easier for real programmers to write and maintain real programs. That's a sign of maturity, and although I love shiny new baubles, to get real work done I'd rather have maturity (which I why I'd avoided Java for so long).
If someone's going to "retire" languages with C-like syntax, they ought to wait until C++ goes into a dark alley and make sure it leaves in a body bag.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 22, 2009 6:41 UTC (Sun) by bojan (subscriber, #14302) [Link]
That would bankrupt the hardware companies. No more CPU/memory eating software... Oh, wait - IBM is a hardware company ;-)
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 0:16 UTC (Fri) by jamesh (guest, #1159) [Link]
I guess this would give IBM access to the UNIX license that Sun acquired from SCO ...
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 0:34 UTC (Fri) by giraffedata (subscriber, #1954) [Link] Clearly the market likes the idea of IBM snapping up Sun That's certainly not clear from the facts in the article. All the article says is that the price of Sun stock went up to $8.31/share, which tells us only that the market believes the rumor that IBM will pay $9-10. To see what the market thinks of the combination: Sun added 2.7 billion dollars to its market capitalization the day of the news, while IBM lost 1.3 billion. But the S&P 500 went up 2% that day, and the same factors probably increased both companies values the same amount. So the news apparently added 2.6 billion to Sun and sucked 3.8 from IBM. That tells me that the market thinks Sun and IBM together are worth less than they are apart. Also that Sun is worth less than IBM thinks it is.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 2:10 UTC (Fri) by rahvin (subscriber, #16953) [Link]
The market isn't an entity. There are no conclusions to draw from any movement in stocks. The market as an whole is like the deranged drunk guy that see stumbling around downtown. It can react violently to anything and it can completely ignore rather important events. It also can react for completely no reason. I always think it's humorous when someone tries to project meaning or reasons for market movement. The news media is partly to blame for giving reasons for the market being up or down for the last couple decades when in reality there is often no reason or intelligence involved.
So in summary, the "market" didn't predict anything. You might argue it moved because of that, I might argue its movement was a direct result of people thinking it was moving as a result of the news when it was in fact just a bunch of irrational trading.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 3:50 UTC (Fri) by butlerm (subscriber, #13312) [Link]
The market can certainly react for the wrong reasons, on bad information,
on bad interpretations of good information, according to completely unrelated factors, and so on.
However, the law of large numbers pretty much guarantees that a stock
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 5:15 UTC (Fri) by jordanb (guest, #45668) [Link]
Sure it doesn't move for literally "no reason" but rahvin's point is still valid: the stock market is good evidence that there's little wisdom in crowds. We're about as well off trying to divine useful meaning from chicken guts as we are from its movements.
I, personally, think that IBM will surely smother any invigorating spirit left at Sun. Certainly we're at the stage where there's a question if Sun will still exist in a form more substantial than that of SGI in a few years, so one might say that anything -- even absorption into IBM -- would be welcome if it kept the company a going concern.
I wish IBM would somehow take a non-controlling interest in Sun rather than consuming it outright, although I don't know what thought process at IBM could result in that as they're direct competitors. At any rate I think we're witnessing the passing of a company that might have become one of the most potent promoters of Free Software in the corporate world.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 5:53 UTC (Fri) by kune (guest, #172) [Link]
If I read anything from the raise of Sun stock it's the fact that the market didn't predict IBM talking to Sun. It shows that the market participants don't know more than anybody else.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 20, 2009 16:07 UTC (Fri) by giraffedata (subscriber, #1954) [Link] the stock market is good evidence that there's little wisdom in crowds. We're about as well off trying to divine useful meaning from chicken guts as we are from its movements. A good point, but not really relevant to this article or my comment on it. The article doesn't use the stock price changes as a predictor of the future or evidence that the merger would be good. All it says, by way of anthropomorphising the stock market, is that the price changes tell us people buying and selling stocks tend to like the idea of the merger. (And I said they tell us the opposite). I wish IBM would somehow take a non-controlling interest in Sun rather than consuming it outright, although I don't know what thought process at IBM could result in that as they're direct competitors. I don't think the competition would affect an IBM decision to buy a non-controlling interest. Even a non-controlling interest reduces competition (if Sun takes market share from IBM today, IBM loses all that profit; with a 49% interest, IBM loses only 51% of the profit). The thought process at IBM that would bring about a noncontrolling interest purchase would have to be thoughts that IBM doesn't have enough cash to risk on the whole company. There's no other way IBM leaders would voluntarily tie their own hands. And in addition, IBM leaders would have to believe that Sun will suddenly start being a lot more profitable than it has been recently.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 21, 2009 0:04 UTC (Sat) by jd (guest, #26381) [Link]
Yes and no. The stock market doesn't care if something's good or bad for the company, the products or the consumer. However, like all good Ferengi, they can smell where's there's profit. Stock prices are based solely on what people think other people think the company is worth, with an eye to buying (or selling) before the other guy does.
What this means is not that the company is worth $8/$9 a share, or that anyone else believes it is. What it means is that enough people think that other people think that it's worth that, which could just mean that someone started a rumour that IBM would pay, oh $10 a share and the rest of the trading floor bought into it. It could also mean someone did the analysis and came up with such a number. It could mean almost anything, making almost any conclusion impossible.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 21, 2009 3:04 UTC (Sat) by giraffedata (subscriber, #1954) [Link] You're talking about speculation, which is only a small part of the pricing of stocks in the stock market. Historically, when speculation has become the major factor in pricing, the market crumbles like a pyramid scheme. You can't have prices based on nothing but themselves. Prices of stocks are always anchored by the everpresent possibility that the corporation will liquidate, and thus are based to a large degree on actual value. Lots of professional investors tell you they invest based on "value," "quality," or "fundamentals," which means they spend their days looking at business documents and make an offer to buy or sell based solely on what they find there.
... which could just mean that someone started a rumour that IBM would pay, oh $10 a share and the rest of the trading floor bought into it. And that's what we're assuming happened, thus giving us information about what the market (i.e. traders) thinks about a Sun/IBM combination. It gives additional information too, because people would not buy into the rumor unless they believe $10 a share is somehow a plausible amount for someone to pay for Sun's assets. And it disproves the speculation theory too, because speculation would cause IBM stock to go down to about zero as the pessimism about overpaying for Sun feeds on itself. But the price of the stock is propped up by the fact that IBM is actually able to generate wealth. So IBM stock didn't go down by more than the actual amount traders think IBM will lose on the deal.
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 23, 2009 20:29 UTC (Mon) by davecb (subscriber, #1574) [Link]
A colleague noticed that Sun's stock
price has a steady, small saw-tooth pattern, suggesting that someone's buying or selling a lot of stock in small parcels.
This may be a puff-the-stock operation...
--dave (who is undecided on the good/bad)
bad for Lustre? Posted Mar 20, 2009 8:43 UTC (Fri) by PlaguedByPenguins (subscriber, #3577) [Link]
It's not all sparkles and sunshine... What happens to Lustre in this deal?
Lustre is a high performance open GPL filesystem maintained by Sun that competes directly with IBM's proprietary closed GPFS. IBM probably makes a lot more money with GPFS licenses than they could ever do with Lustre support, so it seems likely they'll kill off Lustre. That would be big loss for the HPC Linux community...
bad for Lustre? Posted Mar 20, 2009 11:06 UTC (Fri) by jengelh (subscriber, #33263) [Link]
If they kill it, someone will fork it, for it is GPL.
Monetizing Java Posted Mar 20, 2009 19:19 UTC (Fri) by man_ls (subscriber, #15091) [Link] Sun owns some very valuable software properties including Java, MySQL and VirtualBox, items that IBM could well monetise if it could get its hands on them.Monetize Java? IBM is already doing very well selling Java enterprise software. They offer middleware WAS servers, and a host of products on top: WID, PS... The author in passing says something about Websphere, but he gives the impression that it is a single product. In fact WAS (Websphere Application Server) is a huge franchise growing bigger all the time. Add to this the Rational family, also a huge franchise (and in some fronts integrated with Websphere). RAD dev environments have hundreds of components. They are based on Eclipse, but completely proprietary. All these Java-based products are probably selling billions every year. I don't know the figures, but I bet IBM outsells Sun by 10x or 20x, given the latter's puny enterprise offerings. With Java going free I cannot imagine IBM earning much more from it...
It's not the question of "earning much more"... Posted Mar 20, 2009 19:51 UTC (Fri) by khim (subscriber, #9252) [Link] All these Java-based products are probably selling billions every year. Yes - and this is enough money to buy Sun. Face it: if Sun will go bancrupt and will be unable to support Java all these expensive contracts will be reviewed again. How can you offer support for all these "enterprise Java solutions" if you can not guarantee the future for the Java itself? The same is true for Lotus Symphony. Even if IBM will not get any additional money from this deal it'll be worth it as form of protection. The one thing IBM does not need is Sun's hardware business...
A question of divestiture Posted Mar 20, 2009 20:39 UTC (Fri) by man_ls (subscriber, #15091) [Link] True, Sun's hardware business is not exactly flourishing either. IBM did already sell their PC business to Lenovo. If Sun's lines do not offer much to IBM (and they don't seem seem to), it could well mean the end of Niagara and similar efforts. Anyone knowledgeable care to comment?The question remains open for Solaris. I don't know if IBM needs any of that for their AIX business. The author of the article suggests that reunification could happen around Linux, but that is a long shot -- IBM already tried to sell Linux and went back to AIX in horror. Again, insights would be most welcome here.
A question of divestiture Posted Mar 20, 2009 22:13 UTC (Fri) by drag (subscriber, #31333) [Link]
Well there are other companies that have been working with Sun on their CPU business. Fujistu is a big potential customer.
It would be to bad too see it die. Having open source CPU designs is quite impressive coming from a company like Sun Microsystems. (see also OpenSparc)
A question of divestiture Posted Mar 21, 2009 13:03 UTC (Sat) by Wol (guest, #4433) [Link]
IBM's attitude is "if the customers want to pay for it, we'll do it". They've bought quite a lot of stuff like that for the maintenance stream. And I think they've often been pleasantly surprised how much customers are prepared to pay.
I don't think "they went back to AIX in horror". Their customers preferred to stick with the devil they knew, and not migrate to Linux. IBM couldn't care so long as they're making money.
Cheers,
A question of divestiture Posted Mar 21, 2009 17:56 UTC (Sat) by nix (subscriber, #2304) [Link]
Ironic, really. Everyone else who looks at AIX goes back to Linux in
horror. ;)
A question of sticky devils Posted Mar 22, 2009 23:30 UTC (Sun) by xoddam (subscriber, #2322) [Link]
In both cases, it's merely 'better the devil you know'. In my experience (very limited in the case of AIX) the similarities are more daemonic than the differences ;-)
Sun rises on talk of IBM deal. Good for Linux? (Tectonic) Posted Mar 21, 2009 7:22 UTC (Sat) by stuart2048 (subscriber, #6241) [Link]
I'm cracking a beer. Cheers to any fellow ex or current Sun employees out there.
Scratch that... Posted Apr 6, 2009 15:15 UTC (Mon) by pr1268 (subscriber, #24648) [Link] Looks as if IBM is Abandoning the deal. Oh, well...
|
|||||||||||||