The arguments for vendor support of Linux are simple economics. In HPC this usually means that a vendor has a reasonable expectation of return on their investment if they support Linux. This is a valid view if you have something of value that people want at a price they are willing to pay.
The arguments for computer vendor support of Linux are again, simple economics. If Linux is driving business, you expect a vendor to pay attention. Revenue trumps ideology.
Well, ZDnet has an interesting set of comments from Dell.
Speaking to ZDNet.co.uk at the Gartner Symposium/ITxpo in Orlando on Thursday, Dell’s chief executive officer said his company has seen Linux uptake for servers increase faster than Windows server products, despite Microsoft’s claims.
“On the server side Linux continues to grow nicely, a bit faster than Windows,” said Dell. “We’re seeing a move to Linux in critical applications, and Linux migration has not slowed down.”
This is interesting. This suggests that a) the IDC numbers may be believable as a lower estimate of adoption, b) other vendors are noticing similar things, c) there are revenue impacts of these migrations. Vendors pay attention to demand, they have to. Both hardware and software vendors have to.
In the past, bundling deals usually arrived with a reduced price per shipped box, and some sort of marketing money. The former effectively lowered the cost of the bundling operation, and the latter was (in theory) to help the company market the product. Unfortunately this tends to discriminate against the smaller players who can’t get the attention of the software vendor. Their costs are effectively higher for the same product. So if the demand is strong, they are at a profit disadvantage to other vendors. In the case of Linux, the demand for it preceded bundling. Vendors realized that end users were either buying the bundled machines and wiping them (hence the lower bound of the IDC numbers), or demanding unbundled machines. This happens in enough cases, and the vendor realizes it is leaving money on the table. There is value in them thar Linux hills. This demand could be monetized.
And this is what is driving pretty much all the hardware vendors out there to supply either unbundled or Linux offerings. Now some of the vendors are making a huge mistake, and consigning their pre-loaded Linux onto a small fixed number of models, which might not have the features that are needed.
Design something to fail and it often will.
I am typing this on a Dell Precision M4300 workstation (it is a very nice laptop!!!) purchased in August 2007. I had two choices, XP or Vista. No third choice. I purchased this with XP. I run XP on this once every 2 weeks or so, mostly to update the anti-virus bits. It runs Linux pretty much continuously. As soon as I can, I will completely virtualize the windows so that I run it, when I need it, in a VMware session.
Yet, despite this, Dell, IDC, and Microsoft very likely counted this workstation as a win in the windows column.
Remember that when you look at the IDC numbers. And the Dell sales figures. Linux is growing hard and fast, outpacing windows according to them. On the server side. How many laptop owners, desktop owners, etc are wiping windows to put Linux on them? I don’t think it is a microscopic or miniscule number.
Curiously, this phenomenon has shown up in HPC as well. The NCSA Lincoln cluster seems to have disappeared from NCSA. Whether it has been remissioned, or just never rebooted into Windows, well, this is not known by me. What I do notice is that, following the link above, this highly touted resource does not appear to be available at NCSA. No explanation as to where it has gone is easily visible, though I do invite someone in the know to comment.
It is visible in the June 2006 Top 500 list as entry 130. Its specifications are Dell PowerEdge 1855 blades at 3.2 GHz, with Infiniband, running Windows. The press releases suggest it is dual boot. In November 2006, that cluster is no longer in the list. In June 2007 it is missing as well.
You don’t go from 132 to off list that quickly, unless you were pulled off, decommissioned, …
More to the point, in June 2007, we see the Microsoft Windows cluster group seems to have just purchased a Dell Poweredge 1955 based cluster. It turns out you can run the 1955 blades in the same chassis as the 1855 (as far I have have read). That and the Abe cluster appears to be a Poweredge 1955 running Linux.
So I have to guess that either a) NCSA absorbed Lincoln into the Abe cluster, got a retrofit or new set of blades from Dell, or b) they sold the Lincoln cluster, lock stock, and windows boot to Microsoft, whom upgraded it, or c) they decommissioned it.
I think “c” is highly unlikely. This would cause problems. Even “b” would be unlikely, though less so. Microsoft likely bought their own. Which leaves “a”. And yes, this means (quite likely) that it was absorbed and not rebooted.
Well, ok, there could be a “d”. and that would be someone just forgot, and didn’t get the submission in. Twice. After great fanfare … Ok, that isn’t so likely.
But hopefully you get my point. Which is that this much heralded cluster is not in fact easily visible to the Google panopticon, or on the NCSA pages. Which suggests something actively took it off. By absorption (most likely), or other methods. And it probably isn’t running windows any more, unless Microsoft is running it.
There are currently 2 windows clusters on the Top500, and one of them is owned by Microsoft. The third appears to have been reloaded. With Linux.