Return to the Press page

LWN.net Weekly Edition for May 16, 2013

PostgreSQL 9.3 beta: Federated databases and more

LWN.net Weekly Edition for May 9, 2013

Weekly edition	Kernel	Security	Distributions	Contact Us	Search
Archives	Calendar	Subscribe	Write for LWN	LWN.net FAQ	Sponsors

Dell's Linux desktop line keeps expanding (DesktopLinux)

[Posted July 8, 2007 by ris]

DesktopLinux reports that Dell is expanding its consumer Linux line. "When Dell first announced that it would be releasing Ubuntu Linux-powered consumer desktops and laptops, some people saw it as more of a stunt than a serious business move. They were wrong. Dell has already expanded its consumer Linux line, and now it has announced that it will soon be offering Ubuntu Linux systems outside of the United States and for new businesses."

(Log in to post comments)

Posted Jul 8, 2007 20:46 UTC (Sun) by sbishop (guest, #33061) [Link]

I'm glad to hear that sales are strong, even better than expected. Perhaps some of Dell's Linux-specific engineering resources could be applied to the s2r and hibernate issues being dealt with right now.

I'm a little frustrated with the way those twin challenges have been approached on the LKML for a while now, especially the last few days. There seems to be two kinds of people involved: those that think it's easy, and those that are actually involved. Somehow it's become accepted that anyone who works on the PM code is an idiot, and that if we tell them that they're idiots often enough, the PM code will fix itself. I suspect that this is due to Linus's rants from about a year ago, along those same lines...

Anyway, to bring this back on topic, perhaps Dell could sucker an engineer or two into helping out with this. Myself, I'm saving up my pennies for one of their laptops. :)

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 8, 2007 21:08 UTC (Sun) by drag (subscriber, #31333) [Link]

Yep.

That Santa Rosa laptop, the 1420n is about what I've been waiting for. Dual Core, X3100 graphics, and option bluetooth. As well as that probably-very-glorious 1440x900 widescreen. To bad they don't have the 802.11N support or camera, but oh well. I suppose I could always upgrade later on when I start seeing 'N' networks. Ebay is a nice place sometimes.

A friend told me that the local Best Buy (I think that is what he said) had Ubuntu/Dell computers in it's store. That suprised the heck out of me. This is midwest USA... This is like 99% Microsoft-land. I don't even think they offer Apple computers there, although it's been a while since I visited the place.

I bet you that when Dell starts with it's small business Linux offerings then sales will at least double as Linux geeks convince their employeers to upgrade their workstations.

If they start offering this stuff in Europe I bet sales will double again, at least.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 12, 2007 22:38 UTC (Thu) by kamil (subscriber, #3802) [Link]

With respect to 802.11n, I have an older (~6 months) Dell laptop, and I verified that it *does* have the third antenna connector necessary for 802.11n (it's currently not connected anywhere, just dangling there near the mini-PCI slot). I'd expect things to be the same with this new model. So indeed, it should be possible to upgrade the mini-PCI WiFi NIC to 802.11n, once those become easily available on the market.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 8, 2007 22:49 UTC (Sun) by dlang (✭ supporter ✭, #313) [Link]

go back through the mailing list about two weeks and you will see some other major rants from Linus

the good news is that there is a little bit of change on the part of the people working on this now. however they are still mixing up the end goal with their current implementation.

the problem isn't that people see the issue as easy, it's that they see the current approach as broken, and broken in ways that can't be fixed (even the people working on the process say that they can't make it reliable without 'fixing' 90% of the kernel). they then suggest other approaches that would be easier, but becouse they are different then the current approaches they aren't considered by the people with large investments in the current approach

it may end up that the final solution ends up breaking the old 'suspend' functionality into three different chunks.

1. suspend to ram

this doesn't need to freeze processes at all (the biggest problem with the current approach), just stop scheduling things, put hardware to sleep, wait until wakeup, restore hardware that was put to sleep, and start scheduling again

2. suspend processes to disk

checkpoint processes (and process groups) on the system, write them out to disk, shutdown the system. when the system is booted read in the processes as if they were migrating to a new machine and restore operation. this would even allow kernel upgrades during this process. this may involve virtualization support, especially on the restore

3. suspend system snapshot to disk

this is what the current suspend to disk tries to do. it takes a snapshot of the entire system, including caches and kernel internals, and then writes that out to disk. the advantage of this over #2 is that the system restarts with all it's caches intact, the disadvantage is that the system spends a lot of I/O saving and restoring these caches.

the problem with #3 is that it needs to stop everything so that the system remains consistant, but then it needs to do some things to write the data out to disk. these two requirements are in direct conflict and is why the current freezer runs into so many problems.

personally I have doubts if #3 can ever be made to work well (short of running everything in a sandbox so that there is a 'hypervisor' layer that's outside of the running system to do the work), but right now the problems with makeing #3 work have prevented the much easier #1 from being reliable, and that's what's frustrating people.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 8, 2007 23:25 UTC (Sun) by NightMonkey (subscriber, #23051) [Link]

Geez. I'm not really that familiar with the debate over mainline kernel inclusion, but I just have to say that I've been using TuxOnIce (formerly Suspend2) across two laptops (IBM 600X and now a Fujitsu P7120D) and a Dell desktop at work for about 2 years now for suspend-to-disk, and it's been a wonderful experience. It Just Works(tm). I use the suspend2-sources kernel patchset available in Gentoo, and I've had zero problems, and with 2.6.20+ the speed of the suspend/resume has increased.

Documentation is good, and development is active. http://www.tuxonice.net/

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 16:01 UTC (Mon) by elanthis (guest, #6227) [Link]

You, along with many other people, need to learn that anecdotal evidence is worthless. It works for you. Fan-fucking-tastic. It doesn't work for a lot of other people. If it doesn't work for everyone, it's still broken. You just happen to be lucky and fall into the range of users for which their specific combination of software, hardware, and work habits allow everything to work.

Coders will appreciate this analogy. There was once a novice programmer working on network code for the first time. He was reading lines from a telnet session using a single recv() call, and it always worked for him. Just so happened that all the test users were using clients that sent lines as a single packet and never sent lines long enough that needed to be broken up. The coder refused to accept the explanation that his code was broken because it didn't handle the case when the bytes received by recv() would contain multiple lines or partial lines of input. He "knew" that his code worked perfectly and that there wasn't a problem in the world. Or, at least, it did during testing. Oops.

Anecdotal evidence is worthless. Just because it works for you doesn't mean that it isn't horrendously broken.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 17:21 UTC (Mon) by bronson (subscriber, #4806) [Link]

No, anecdotal evidence isn't worthless. If a program works for other peple out there, that means that there's some chance that it will work for me too. That's good information, especially if I happen to own a 600X or P7120D.

You, along with many other people, need to learn that sweeping generalizations are always bad.

(irony intended :)

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 20:10 UTC (Mon) by zlynx (subscriber, #2285) [Link]

Using your own example, if the program using the single recv() does not work, the program written by the sender is not broken, *even* *if* it does not work for that one guy.

If someone's hardware has a buggy ACPI and a fully ACPI standard compliant Linux doesn't work on it, that BIOS is broken. Linux is not.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 22:03 UTC (Mon) by NightMonkey (subscriber, #23051) [Link]

Sure, anecdotal evidence is pretty useless in Nuclear Physics ("I swear that on my way back from the lunchroom, I tripped over a quark!"). But no, I won't learn your "lesson" - anecdotal evidence has a place, and this was it.

I wasn't just "lucky" to use Suspend2/TuxOnIce. I read the docs, I read user's reports and reviews (both good and bad), and made a decision to use it. In other words, it was intention, not luck, which made it work for me.

So, do you have evidence that TuxOnIce is "broken for everyone", or was this just an exercise in hearing yourself pontificate?

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 8, 2007 23:54 UTC (Sun) by sbishop (guest, #33061) [Link]

Thank you very much for the summary. This can be a difficult discussion
to follow sometimes.

Am I correct in understanding that both options 2 and 3 require that
userspace be stopped between when the snapshot is taken and the system is
restored? Without stopping userspace, all kinds of troublesome things
could happen--all of them filesystem related, that I can think of--such
as filestamps, directory contents, and programs/libraries changing, which
would confuse a restored system. This, in turn, would mean that saving a
snapshot to a FUSE mount would be impossible. Is that right?

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 0:43 UTC (Mon) by dlang (✭ supporter ✭, #313) [Link]

yes, with options #2 and #3 userspace is frozen from the time you start the suspend so you can't do anything that requires your main userspace

however there are three things that can be done to give you a useable userspace to work with.

1. only freeze some of the userspace (this is the approach that has been tried in the past)

this can be horribly complicated (trying to figure out what to freeze and what not to)

2. use kexec to start a new kernel (with it's own userspace)

this eats up a chunk of ram permanently (8-64M per Pavel)

3. use a hypervisor and run the entire 'normal' userspace as a virtual system and then use the hypervisor's 'userspace' to do the write out.

this is a speed penalty at all times that will depend on the type of virtualization that you do. (5-15% penalty is the numbers that Pavel threw out)

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 20:01 UTC (Mon) by skitching (subscriber, #36856) [Link]

Yes, I saw those comments re virtualisation and kexec.

The existing kexec implementation reserves memory permanently because it's used for dumps when the primary kernel has an oops, yes? Obviously in that case a fresh kernel cannot be read from disk and set up because nothing can be trusted to work - drivers, memory allocation, etc. so it needs to be loaded early and hang around *in ram* just in case it's needed. However in the case of suspend/hibernate, why isn't it possible to set up the alternate kernel when needed then jump to it? If so, then the complaint about permanently losing MBs of ram to the kexec kernel isn't relevant..

Similarly for virtualisation, it's silly to run linux in a vm just to have suspend. But if virtualisation helps, then can a running non-vm kernel be "virtualised" on demand?

Personally I don't understand how kexec would help anyway. As I understand it, the problem is that user-mode helpers (FUSE, firmware loaders, etc) can be needed by the kernel in order to process in-kernel data structures during the suspend. But in that case, the new kernel will (a) not have knowledge of that kernel data to be processed, and (b) will not have the appropriate user-mode stuff set up. However I really don't have much of a clue about this...

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 0:44 UTC (Mon) by dlang (✭ supporter ✭, #313) [Link]

here is a post from the list about the freezer that appeard shortly after I made the comment above

> Freezer is not needed for snapshot -- it is needed so that we can
> write out the snapshot to disk without the need for special
> drivers/block/simple-ide-for-suspend.c. (We are doing snapshot, then
> write to disk from userland code in uswsusp).

Yes.

BTW, this patch:

http://www.sisk.pl/kernel/hibernation_and_suspend/2.6.22-...

that's queued up in -mm contains a freezer documentation update, in which the
reasons of using it, as well as its limitations, are described.

To summarize what was previously said in this thread:

* Apparently, we agree that the freezer is _generally_ not needed for suspend
(ie. any transition to a system sleep state other than hibernation), but some
of us (eg. me) think that it wouldn't be reasonable to drop the freezer from
the suspend code path _right_ _now_ .

* Some of us, including you, Nigel and me, think that the freezer is needed
for hibernation (please see the document in the patch above for details).
In the (very) long run this might be avoided too, but (IMO) certainly not at
this point.

* We seem to agree that in order to remove the freezer from the suspend code
path some work needs to be done on device drivers, driver midlayers and the
PM core. We also need to do some work on the PM core in order to introduce
a separate hibernation framework and IMO it would be reasonable to
synchronize these efforts.

* We are now to decide what to do so that the freezer can be safely removed
from the suspend code path and how to integrate that change with the
hibernation code path (if possible and reasonable).

* The freezer vs FUSE issue that started this thread remains unresolved, so
it would be desirable to provide a short-term fix (need not be very nice).

Greetings,
Rafael

dumb question

Posted Jul 9, 2007 15:33 UTC (Mon) by ccyoung (guest, #16340) [Link]

would it be possible to do a little bit like databases, that is, suspend the snapshot and, as new events (transactions) come in, make a log of them and re-apply them once the snapshot has been restored?

dumb question

Posted Jul 9, 2007 15:51 UTC (Mon) by dlang (✭ supporter ✭, #313) [Link]

the trouble is that what you are talking about for events are memory writes, etc, recording all the effects of these (especially the side effects of writing to hardware) is not practical.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 0:56 UTC (Mon) by drag (subscriber, #31333) [Link]

Well I had a Ibook running Linux for a couple years. It did suspend-to-ram perfectly.

Suspend-to-RAM is going to be the most important thing for desktop use. Because it's fast and very convient. Because it's fast then it makes it more usefull.

Intel's next-generation higher end laptops are going to have battery lives that measure 12-17 hours uptime. If you use one of those things at work, as a regular user, only going to use it.. say.. 20-30% of the time. The rest of the time your running around, taking care of business with people, talking on phones, on a desktop, or whatever.

With 10 hours uptime, and 20% duty cycle it would be pretty easy to acheive a full 40 hour work week of use out of that laptop on a single charge.

If you need the mobility then this makes these laptops VERY usefull. On a good setup it will turn on and shutdown faster then it takes to open or close the lid on the thing... Which is a natural thing for people to do. Close it so people can't see what your doing while your gone, fold it under your arm when you move.

Then you have Linux smartphones, PDAs, things like the folio, which benifit heavily from this sort of thing.

Beleive me, after a short time of having something that works instantly, it's much more usefull and it's addictive. Times were kernel changes or whatnot caused it to stop working for me made the device much less convient and it sat shutoff a lot more.

For other uses, like stationary desktops or servers then option #2 is the best, I figure.

With option #3 the hardware has to remain in a constant state so that when everything wakes up the kernel won't go into a funk about it and start panicing or whatnot. But with #2 the kernel you will be running can gracefully deal with any sort of hardware changes or whatnot.

So this will potentially provide a "poor man's hotswap" for low-end Linux servers and workstations. Change out disks (with a FS image transfered over), change power supplies, networking cards, disk controllers, add new ram, or replace old bad ram. Add CPUS, replace bad cpus. etc etc.

Say you had this suspend stuff dead-on reliable. Say your getting warnings about your ECC RAM. So you walk up to the server, press the button on the front. Wait for the light to go out, pull it out, flip the cover up, pull the old ram, put in the new stuff, close the cover, shove it back into the rack and press the button and walk away...

If your quick you'd probably be able to get it done and the machine back up before it's TCP sessions time out. Almost hotswap, from the point of view of end users and aviability it's damn close.

Plus if it ties into all the virtualization and clustering work that is getting done then your killing 2 birds with one stone. Always nice.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 6:20 UTC (Mon) by jel (guest, #38548) [Link]

Unfortunely, "outside of the United States" still seems to mean, "in a few
selected areas". Likewise, "make ubuntu available" seems to mean, "on the
particular machines we decide you can have a choice with". You can still
go onto the Dell site and struggle to find a non-windows machine.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 7:43 UTC (Mon) by tuna (guest, #44480) [Link]

Well, since certain HW makers do not support (the kernel) Linux, it would be quite stupid of Dell to put Ubuntu on them since the user will have much trouble getting the computer to work.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 17:25 UTC (Mon) by jel (guest, #38548) [Link]

Well, no; it's quite stupid of Dell to use hardware that's not well
supported to begin with. Fixing their systems rather than working their
butts off on explaining which ones aren't broken would make a lot more
sense.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 19:08 UTC (Mon) by khim (subscriber, #9252) [Link]

Well, no; it's quite stupid of Dell to use hardware that's not well supported to begin with.

If your goal is to create mega-super-cool system, then it's stupid, sure. You can find list of companies who tried to do it, did that and then went bankrupt easily enough. If your goal is to make profit - it's not stupid at all. And surprisingly enough Dell's goal is profit not some ideals. You take the cheapest good working component and use that. If it does not work for few percents of your userbase who like Linux... tough: they can buy another system with supported hardware.

Fixing their systems rather than working their butts off on explaining which ones aren't broken would make a lot more sense.

Will it truly be cheaper ? Can you prove it ? I sincerely doubt it: Linux-friendly hardware makers are usually more expensive and while difference can be small... if you'll multiply it by millions... it makes absolutely no sense whatsoever to put Linux-friendly HW in Windows system...

When/if another manufacturer will offer better deal for Linux-friendly hardware - we can switch to it and say: "Dell, you are falling behind". So far there are none: either companies are small and [relatively] expensive or their "Linux-friendly offers" are actually worse in real life then Dell's (full of binary blobs, etc)...

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 20:05 UTC (Mon) by sbergman27 (guest, #10767) [Link]

"""
Well, no; it's quite stupid of Dell to use hardware that's not well
supported to begin with.
"""

All of the hardware that Dell sells *is* well supported. By Windows, the operating system that "Dell Recommends" and sells most.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 21:04 UTC (Mon) by Los__D (guest, #15263) [Link]

- And most of it is quite well supported in Linux as well, just not _perfectly_ supported, which is what is needed for Dell to sell as a supported system

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 9, 2007 22:13 UTC (Mon) by sbergman27 (guest, #10767) [Link]

And continuing that line of thought, Dell has the clout to ask their suppliers to please address the matter and be listened to.

But such things do take time.

I'm quite content with the rate at which this is proceeding. Best to keep their standards high and make certain that their customers have a good Linux experience.

Remember that poor execution of this rollout on the part of Dell could hurt Linux's perception by consumers for years.

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 10, 2007 1:00 UTC (Tue) by drag (subscriber, #31333) [Link]

I don't think that it makes sense for Dell to offer Linux on all their stuff.. but it won't make sense for Dell to offer Windows on all their stuff either.

Linux has this advantage over Windows you see.. I don't see many people talking about it, so I'll mention it. If I am off-base here, let me know.

Right now one of the semi-myths going around is that Windows has better hardware support because of it's monoculture and it's mostly-stable ABI. Drivers only need to be developed once, were Linux drivers need to be re-engineered on a pretty much ongoing basis. Or something like that.

Well the way I see it one of the major reasons people have the perception of the superiority of Windows hardware compatability is that Windows tends to 'just work'. Well I figure that it doesn't 'just work' for anybody realy. Driver development for Windows can be very difficult when compared to Linux.

First assumption....

The thing is that if you go into any big hardware manufacturer, say somebody that makes motherboards, your going to find rooms full of Windows computers for testing compatability with various third party drivers and hardware. Test, test, test. They will change the drivers, change how the hardware works in order to make it work with Windows. Testing compatability with video cards, cpus, memory, disks, controllers, etc etc. Everything, trying all sorts of different combinations.

So it seems to me that Windows doesn't 'just work'. People all over, and not just Microsoft, spend a LOT money on making Windows 'just work'. A lot of the development stuff that Windows needs is paid for by people who make no money from Windows sales. That's money out of their pocket being spent to make Windows work. Millions of dollars being spent.

Second assumption.....

If Dell's Linux ambitions take off then in the future they will probably have the Windows and Linux product lines be mostly the same hardware.

For example the E1505 Windows machine vs. E1505 N Linux machine. The Ubuntu machine has Intel graphics with Nvidia optionally. The Windows machine uses ATI graphics. In the future it makes sense that Dell wouldn't bother buying ATI stuff and just have the same laptop with Windows as with Linux. Why have 2 product lines? In Windows-land there doesn't seem to be much compelling reason to choose ATI over Nvidia, maybe a couple bucks or something like that. Not until you get into the high-end gaming/consumer stuff then it switches back and forth.

So in this manner OEMs selling Linux on consumer machines will cause Linux to have a disporpotionate impact on the actual hardware manufacturers. Linux compatability may make or break some very large contracts even if only 2-3% of the hardware being sold will actually end up in Linux computers. This will convince them to spend more money and effort in supporting Linux.

Conclusion....

Once hardware manufacturers spend the money on testing (and sending patches and bug reports) for Linux they will realise that Linux support costs substantionally less then Windows support. Also it would allow them to make more specialized and optimized hardware for special tasks since Linux is very portable to legacy-free platforms and Windows is not.

This means that if they target Linux more then they target Windows this makes more profits for them. They can make the drivers cheaper, they can make the hardware with less expenses for themselves. They make more money even if the hardware actually ends up cheaper for end users. This means that they will advertise Linux support and help promote it in a effort to have more people adopt it, making their lives easier.

Any sort of traction that Windows loses in terms of hardware support and market share, it's going to be very difficult for it to get back.

Of course if, and only if this experiment works out for Dell...

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 14, 2007 22:46 UTC (Sat) by leoc (subscriber, #39773) [Link]

What is the quality of a Dell like compared to a "Lenovo" Thinkpad these days? I was thinking of picking up a T61 widescreen with very similar specs to the Dell but the Dell is $200 cheaper, has more memory and more disk, and doesn't come with a Windows tax. However I've seen Dell laptops in the past and have been unimpressed by the keyboard feel and flimsy feeling construction. Have they gotten any better?

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 16, 2007 6:07 UTC (Mon) by bronson (subscriber, #4806) [Link]

I've got bad news coming from the other direction... The keyboards on the T60, T61, and X60 tablet all feel rather stiff and cheap to me, kind of like the $7.00 keyboards at the computer fair. While the keyboard on my T42p is as close to perfection as I've seen, all the new stuff is pretty mediocre.

Lenovo also crammed a Windows key onto the X60t, making the lower left part of an already small keyboard positively anorexic. They made it really hard to hit the right key without looking.

Thinkpads have lost their edge... It wouldn't surprise me at all if the Dell felt better. *sigh*

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 15, 2007 19:08 UTC (Sun) by mikov (subscriber, #33179) [Link]

Out of curiosity, has anybody here bought a Dell+Linux computer ?

I bet that most people would still buy the computer with Windows not Linux, knowing that it is possible to install Ubuntu on it because Dell provides official instructions, drivers, etc. After all it is very practical to have a legal Windows copy to dual boot into.

That is definitely what I would do (btw, my current laptop is Dell and runs Ubuntu great, but I occasionally need to use Windows as well)

So, I think that what Dell may be seeing is increase in their _Windows_ computers sales. Can they really quantify how much offering Linux is helping them ?

Dell's Linux desktop line keeps expanding (DesktopLinux)

Posted Jul 16, 2007 5:55 UTC (Mon) by bronson (subscriber, #4806) [Link]

Yes, two desktops for work. They're darn good machines for the price and it's SO nice just to plug it in and have everything work instantly. If no local white box guys can give me faster hardware for a similar price, I'll buy six more in August.

All the laptops I order, alas, still run Windows. And it's for the reason you state: dual booting is very important for that one time you get an oddball powerpoint deck or Excel spreadsheet. You don't exactly want to waste your VCs' time telling them to re-send in a more free format...