What are we to make of Sun Microsystems' virtualization effort? In the positives camp, we have a major vendor trying to build a real competitor to VMware. That's probably a good thing for the virtualization market since the likes of HP, IBM and Dell appear content to more or less do what VMware, Citrix and Microsoft tell them. …
Sun's data center completeness
I've been watching OpenSolaris recently.
Sun'll be the only vendor that can virtualize, thin-provision and serve (via FC, iSCSI and iSER, once ComSTAR is finished) the block storage (Volume Manager) and filesystem (ZFS), the OS (Containers), the hardware (Xen & Logical Domains), and the network (802.1q). The end-to-end management tools are sub-par compared to the technology but if Sun can get that act together, Solaris will once again be the most rounded-out data center OS.
JG wrote "Sun'll be the only vendor that can virtualize, thin-provision and serve..."
Sorry, all the stuff you've listed (afaik) can already be done today by IBM's AIX 6 OS, especially on latest kit. And it's a very powerful and complete feature set that the wizards in Austin have provided.
Unless you're talking purely about the x86 space, in which case I'll agree Sun's stuff is coming together pretty nicely.
And before you mark me down as a Sun hater - I've got Solaris running on the laptop I'm using at the moment, albeit running under VMware Server 2.0 Beta. ;)
ITRocks .... Suck IT and See .... Sock IT with C++++S*
"Sun and Microsoft will support each others' systems even when they're running virtualized code. "
Hmmm? AI Sun Networks Driver for Vista Operating Systems?
* Code 4 Change 2 Civil CyberSpaces ..... for Virtually Real Realities, you understand.
I think it's a little unfair to label Sun as coming late to the mainstream virtualisation game. After all, Sun led the virtualisation field in the UNIX space for quite some time and were thought leaders here for a good part of that. True, their x64 story has oscillated badly over the last decade or two (I remember their first painful steps in that space only too well!) but considering this has only really come to fruition in the last couple of years I don't think we can really put them down for it.
One thing they have always struggled with is getting the best out of their acquisitions (need say more than highground!) but I tihnk their x64 virtualisation solutions also shows they have learned lessons here.
Let's just pray we do see valid and strong competition to VMware from Sun - VMware are evolving from strength to strength (especially in the virtual desktop space) but I would hate to see EMC form a stranglehold on the market. Long term choice is good for all of us.
RE: Joshua Goodall
We've already nicknamed it "OopsCenter"
Thanks for the feature sell, Joshua. I'll make sure we take all our existing virtualised environments on hp-ux, Windows and Linux out on Sun's promise that they may have a feature some time in the future which we seem to be provisioning and virtualising without just fine, thanks! Hadn't you better wait until NetApp stop kicking Sun around a courtroom over WAFL/ZFS before you accept any more of their promises?
"The pub calls...." raincoat, please!
More like early, then late...
Sun was one of the first Unix vendors to approach virtualization at all, mainly based on the E10k, domains, and early versions of Solaris Resource Manager in the mid-late 90's. These efforts sold a lot of hardware, but weren't sophisticated enough to allow most customers to really get much benefit.
These initiatives lost steam in the early 2000's, as Sun had bigger problems on their mind. The also attacked the Unix virtualization market with o/s virtualization (containers) rather than with a hypervisor. IBM decided to go with the hypervisor approach first and, along with the addition of a lot of other feature/functionality, gained the lead in the Unix virtualization race. They're still ahead now, IMO. For some reason, HP seems to always get left out of this conversation, even though they have much of the same virtualization mechanisms as Sun/IBM, but just haven't done a good job of marketing them.
On the x86 virt front, Sun does have a pretty good story, but they need to get their hypervisor out the door pronto, also the pricing/support needs to be solid. As for buying SWsoft/Parallels, it would have been a good idea a year ago, but it would take some time to get Solaris support into the products - time that Sun might not have now. I expect that Microsoft and other players might be sniffing around Parallels now...
@AnonCoward & Matt Bryant
AnonCoward: AIX has a subset of that list. Try doing thin-provisioning of iSCSI targets and NFS/SMB exports with AIX (which Solaris does instantly, and on a thousands-of-volumes scale), or installing Linux in a WPAR.
Matt Bryant: Check your facts - this isn't vapourware. My group is already using most of these things in limited production via OpenSolaris SE releases. NetApp are just scared - and rightly so, although competing via the courtroom is pretty lame. Given Sun's indemnification of customers, there's no need to wait.
RE: Joshua Goodall and Col. Zander
I think you'll find NetApp aren't scared at all, they're just p*ssed off with Sun pinching their work, then trying to open source it in a bizarre attempt to sidestep litigation. It's the equivalent of a thief breaking into your house and when he's chased down the street, throwing your goods in everyone elses' front gardens under the impression that giving them away for free somehow alleviates guilt! In such a situation would you describe calling the police "lame" too?
Like many Sunshiners you also massively missed the point - most customers just aren't interested into thin clients, they'd rather have full Windows desktops because it makes their workers more productive. And if they want to go thin they want thin Windows clients, usually via VMware, not Solaris ones. Screaming that your poo is better than anyone elses just because the guy next to you is selling hotcakes does not make your poo sellable, it just means your product stinks and the hotcakes will sell. If your poo is beta or prerelease poo (vomit?) it doesn't make it any more attractive to the average hotcakes buyer.
And for the Colonel, I think you'll hp-ux had npars and vpars (hardware and software partitioning) available to buy before Slowaris or AIX had true partitioning.
Ashless - sorry buddy - you messed up!
Ashlee - you don't often get it wrong - but this time you definitely messed up! :( Where?
a) virtualbox.com - Suns recent acquisition and available for immediate download and trial. Please check it out! Users have reported setting up a virtual Windoze XP in under 30 minutes. And it runs ***fast*** - much faster than QEMU. And faster than Suns implementation of Xen. Oh yes - you forgot about the qemu project under OpenSolaris too! :( Strike 2
b) vmware: Solaris and OpenSolaris already runs under vmware - except that the vmware *host* cannot be (currently) Solaris or OpenSolaris. But ... using a Windoze or Linux host, you can run Solaris and OpenSolaris guest VMs.
c) Ultimately its all going to boil down to *speed*, reliability/stability and manageability. If a user discovers (by trial/error) that they can run more vitual instances under OpenSolaris using Virtualbox than they can under vmware - then which VM environment do you think they'll run? Do you really think that the current generation of freetards has any loyalty to vmware - or any other provider of OS implementations? No - they want cheap (free), fast, reliable and easy. And they won't pay vmware a dime if they can get cheap/fast/free from anywhere else!
d) in the enterprise, management is going to be more concerned with manageability that *anything* else. They don't care about cost, care somewhat about speed and care a great deal about reliability and support. These are areas (as you know) where Sun has always succeeded. This is their playground!
Sun is ahead of the game IMHO. Its just that the "game" is currently wide open, and being an early leader like vmware has nothing to do with who/what wins out in the end. In fact, the computer landscape is littered with "early leaders" who did'nt survive in the long run.
BTW: one final point: virtualbox is damn good right now; imagine how much better it'll run with the brilliant minds of the Solaris kernel engineers - you know - people like Bryan Cantrill (DTrace) and Jeff Bonwick (ZFS and the slab memory allocator) - and a cast of dozens of talented (major) players.
Sun is correct with their policy of "no software before its time" with Xen. Everyone is totally pissed when their virtual machine goes "whoops" after investing serious man-hours into a project or task.
Sometimes the turtle wins the race.....
"I think you'll find NetApp aren't scared at all, they're just p*ssed off with Sun pinching their work"
Yawn....the true analogy is about a misfit granted exclusive right to idea that should not have been patentable at all. Now that Netapp finds that the cash flow may dry up slowly with users getting most of the usefulness of WAFL and a lot more for free and on any commodity hardware. The days of milking the cow for ever without feeding it are coming to a close, just watch how these patents get invalidated one after the another. Netapp are scared of competing in the open, they want a sanctuary where no one can come in !
"Like many Sunshiners you also massively missed the point - most customers just aren't interested into thin clients, they'd rather have full Windows desktops because it makes their workers more productive. And if they want to go thin they want thin Windows clients, usually via VMware, not Solaris ones. Screaming that your poo is better than anyone elses just because the guy next to you is selling hotcakes does not make your poo sellable, it just means your product stinks and the hotcakes will sell. If your poo is beta or prerelease poo (vomit?) it doesn't make it any more attractive to the average hotcakes buyer."
That's like howling at the moon. Vast majority of customers, specially big corporations who use windows today on desktops really don't care whether they have windows on their desktop or a thin client as long as the operation, maintenance and administrative costs are low. That's where desktop virtualization delivered over thin client comes handy. Anyone with half a brain will understand that any desktop virtualization strategy, be it Vmware, Citrix, Sun is about delivering the applications into a desktop or thin client, and if these applications are on windows, so be it. There are always users who need the whole desktop power right on the desktop, but there are many other who don't. There is nothing about solaris on any desktop virtualization.
"And for the Colonel, I think you'll hp-ux had npars and vpars (hardware and software partitioning) available to buy before Slowaris or AIX had true partitioning."
Very true, But IBM micro-partitioning being so much better in every way, it just does not matter that HP had npar/vpar for a long time. Only recently HP came up with Integrity VM, which I believe is a VMWare like true virtualization solution for running Windows, Linux and HP-UX on their Itanics. Not to mention HP-UX shipments are so much lower compared to AIX or solaris. With XVM targetting x86 based on Xen para-virtualization approach and supporting Linux and Windows on x86, it's suddenly much bigger a market and quite more interesting.
re: Sun "pinching their work" - get real
Matt - are you serious? Sun did'nt pinch any NetApp IP; instead they rejected NetApps RAID scheme in favor of a technically superior RAID implementation and set of (on disk) consistency checks/balances that NetApp will have to play catchup on to reach parity with. They (Sun) replaced:
1) a closed source (proprietary) software stack (NetApp) with an open stack
2) a custom hardware based implementation with one that'll run on commodity hardware
3) a very good RAID implementation (from NetApp) that relies on non-volitile memory with one that'll run without without custom hardware/non-volitile memory.
4) a custom hardware platform (NetApps) with generic off-the-shelf hardware (anything that'll run Solaris/OpenSolaris).
5) A nasty sales model (NetApp - "let me talk to my manager") with one that allows the user to bring-their-own-hardware or buy anyones commodity hardware.
6) a (NetApp) sales model that says "thou shalt use our over-priced disk drives or else" with one that allows you to use any commodity disk drives - including large capacity and inexpensive SATA drives that you pickup, on sale, at Frys.
7) A reliably error checked system - in terms of data retrieved off the disk with a system that provides much more robust end-to-end error checking. You can totally rely on the data you retrieve from a ZFS based filesystem.
8) An expensive (NetApp) based proprietary hardware storage platform with one based on commodity hardware and disk drives the *kills* NetApp in terms of cost-per-gigabyte.
9) A system that runs fast (no disputing that NetApp performs) with a system that will scale well into the future - as installed processor cores per system and installed RAM per system follows the predicted scaling. A given NetApp system is a case of "what-you-see" is "what-you-get" now and forever. Unless you want to buy NetApps CEO a new yaught and replace your current NetApp platform entirely.
I could go on... But, seriously, would you buy NetApp shares today? You could - but I'd recommend you buy them "short". :) Thats the only way you'll make money on NTAP (already down from ~32 to ~20 in the last 6 months).
Actually, Matt, Windows runs on Solaris xVM right now. I like the poo metaphor, though, nice pro touch there. p.s. Thin Provisioning isn't the same as Thin Clients.
In other news, it turns out that ZFS achieves things like snapshots and writeable volume clones through a completely different mechanism to WAFL. You could even read the published papers for each and discover this for yourself.
After seven years recommending and deploying NetApp to customers, I'm now a proponent of ZFS. And only a year ago we were deploying VMware as the virtualization stack; this new Xen+ZFS solution is much easier to manage and scales further. Make of that what you will.
While his other points can elicit interesting debate, Snafu Two suggestively asked:
"... But, seriously, would you buy NetApp shares today? You could - but I'd recommend you buy them "short". :) Thats the only way you'll make money on NTAP (already down from ~32 to ~20 in the last 6 months)."
Anyone who follows (or even worse owns) Sun stock will see thru this ruse. The facts are that Sun stock has performed just as miserably as NetApp stock during this period.
To make his point, Snafu Two carefully cherry-picked the 6 month high and low for NTAP. On October 23rd, NTAP hit a 6 month closing high of $32.06 USD. On March 31, 2008 NTAP closed at $20.05 USD. That represents a decline of 37.5%. To exaggerate his claim even more, he might have referenced March 19 when NTAP closed at $19.49 USD. That would represent a decline of 39.2% from the October 23rd high. However, if you take the "true" 6 month period of October 1 thru March 31, the decline in NTAP was "only" 24.9%.
So how does the performance of stock in that technically superior, open source, open hardware, sales-friendly, technological powerhouse known as Sun Microsystems (JAVA) compare to NTAP over these same periods?
October 23, 2007 to March 31, 2008 -34.7% -37.5%
October 23, 2007 to March 19, 2008 -34.2% -39.2%
October 1, 2007 to March 31, 2008 -31.8% -24.9%
Someone without an agenda might correctly ask "would you buy NetApp OR SUN shares today?" Neither company has rewarded investors (longs) during the past six months. They have both disappointed their shareholders.
Of course, we might discuss each company's 5-year performance up thru MArch 31st during which JAVA gained 15.5% and the lowly NTAP gained "only" 79.9%.
Enough already on the financial comparisons. The technical discussion is much more interesting and subjective.