Slashdot Mirror
Sun Bare Metal Hypervisors Now GPLv3
ruphus13 writes with some more news for people foretelling the death of VMware. Sun has open sourced their xVM server, their bare-metal hypervisor virtualization solution. What used to once be the cash cow for VMware is now coming under increased threat, and Sun is once again turning to the Open Source community as a weapon. "Sun xVM Server is an outgrowth of the Xen project — which raises the question of why a company would go with Sun's version rather than the Xen one. Apart from its support for SPARC and Solaris (as well as other chips and operating systems), Sun is also building a services and sales organization around a commercial version of xVM server... If you want to kick the tires or cut your costs, you can hop over to xVMServer.org, download the source (GPL 3) and join the community. But Sun is betting that, as deployments move from an initial testing phase to active usage, large organizations will be willing to pay for guaranteed support (starting at $500 per year per physical server)."
$500 bucks a year per physical server is pretty cheap in the grand scheme of things. Basically, you can try out and use it for free as you set the server(s) up, but when you go live, you can have the assurance that proper support brings. Or not. Your choice. Good move on Sun's part.
This guy's the limit!
... why a VM has to "support" a given OS such as Vista or Solaris or Linux?
FTA: "Apart from its support for SPARC and Solaris..."
Surely if these VMs truly are PCs emulated in software with standard emulated devices then surely any OS than runs on the PC architecture and has drivers for these devices will install and run on these VMs regardless?
I'd be leery of any company promising me guaranteed anything. Guaranteed support could mean anything from a full-fledged support staff to an automated phone system designed to loop callers back upon themselves.
Attention all planets of the Solar Federation! We have assumed control! - Neil Peart
vmware does not make its money on bare metal hypervisor. It makes a fortune, and is actually doing pretty good, on enterprise products like vmware infrastructure or virtual desktop environment.
Actually their bare metal hypervisor - ESXi comes for free as well (although not GPLed, but we're not talking about ideology here are we)
They should follow the beer-ware model. They won't get rich, but boy will they have fun!
Please put that under a new license.
More info here
Don't fight for your country, if your country does not fight for you.
I tried this out a month ago on opensolaris, linux, and solaris10 (both on x86). I've historically been a big supporter of Sun, but .. well .. It just 'didnt work' with solaris as the 'guest' OS. The guest would start up, launch X, and freeze. Default options for host&guest&xVM... Not a good start.
http://kenai.com/projects/xvmserver/forums/120-Announcements/topics/59-First-open-source-release-of-xVM-Server?
This release is designed to allow interested parties to view the code - not run it. It will be some time in the future before we have all of the pieces available for you to compile and run your own copy of xVM Server.
But stay tuned, we're getting there :-)
scott
Because the last time I checked, this was x86 only.
One thing people here are looking for in virtualized environments are snapshots and disaster recovery simplicity. All the products handle moving VMs from one box to another differently, and so far I've heard VMWare is much easier than the rest. So it's not just hardware and OS support, it's ease of management -- which is VMWare's strength (though you pay for it).
VMware offers their basic server for free with $350-$450 a year optional support. I don't really see suns offering as much of a blow against them.
Hikery.net - The best hiking site ever. Made by yours truly.
Which has a boatload of problems. The fact is there is enough competition in the market that just being able to be a hypervisor is not enough - you need to measure up and offer proprietary advantages.
The reason this release is not a big deal is that VMWare spanks the performance of every other hypervisor. VMWare ESX networking is magnitudes ahead of every single other competitor in the benchmarks.
Website Hosting
Speaking of Sun, here are some pics of the company's factory in Scotland. If you like servers it's one to check out...
In other words, will this new xVM run unmodified operating systems on ordinary 32-bit hardware that doesn't have hardware VM extensions?
Tired of FB/Google censorship? Visit UNCENSORED!
and didn't put vmware out of business... arguably, sun's hypervisor isn't any different.
The installation of xVM itself on my late-model Dell desktop running a fully updated Windows XP OS but I could never get Ubuntu to install and/or run on three separate attempts. The first time, the Ubuntu install process froze. The second time, it completed but when shutting down to reboot post-install, I got hit with an near-endless stream of error messages and the OS never rebooted. The third attempt also apparently installed but wouldn't boot.
They do claim to support Ubuntu as a guest OS but my experience was a bit different. Your mileage may vary. In any case, I uninstalled it and chalked it up to simply not being ready for prime time.
No OS uses every possible function of the hardware beneath it, only a subset. Hence, each emulator only emulates those functions, vice the entire hardware set. Smart programmers don't write code they don't need.
Adding an OS to an existing emulator may be as simple as adding a few additional functions to extant code. In real life, though, it can require rewriting many modules to account for unique OS specific behavior and parameter passing bugaboos.
Invenio via vel creo
Sounds like a porn magazine for the robot-inclined.
...Do want.
-Sun forks a opensource project - Xen
-Sun continues developing their fork in a propietary way
-Now they release it as opensource! OMG we opensourced it!
Sorry, but this is not interesting. This looks like the typical "I'm going to make my own fork" effort. Sun has probably already lost many of the features being coded in Xen right now just because of the fork.
-IDE drives are not supported as in who uses them in a real server anyway??
-NFS/TCP/IP/ethernet remote storage * CIFS remote storage -- as in that's not enough??
-NICS supporting the Solaris GLDv3 driver specification-- fine
-only MTUs of 1500 bytes are supported (when did you see one smaller recently??)
- Windows Server 2008 logo certified hardware-- that's about all of the servers I know, sadly.
Sun may thwart this one, too, but I'll give them a fighting chance. The model's somewhat sound but I'm eager to see the perf numbers and the real availability costs.
---- Teach Peace. It's Cheaper Than War.
Just a note, you won't ever see unfavorable benchmarks to VMWware, because they explicitly forbid publishing benchmarks without their authorization. Some time back (maybe a couple years now) VMWare published very biased results showing that (well Optimized) VMWare outperformed (Completely unoptimized) Xen. Xen shot back with Apples to Apples numbers where they showed that in a few special cases they were more less equal, but everywhere else Xen killed in the benchmarks. VMWare promptly threatened suit and pointed to their EULA which effectively says: Use of our product guarantees your agreement that you will not publish any comparisons with other products without our consent.
If you want to find the comparisions, google for "xen vs vmware benchmarks redacted" and see if you can find a copy of the unedited results.
B) Eliminate all the stupid users. This is frowned upon by society.
sun's been going on about xVM for god knows how long and it's still "Coming soon" instead of talking they should actually release the product.
I've been tracking xVM for a while now, along with the other major VM players, for my home VM setup. I've downloaded and evaluated ESXi, XenServer Express and Hyper-V. The one difference that xVM will have that the others don't is a web interface for administrating the VMs. All the others require a Windows application, which in turn requires Windows (I haven't tried using Wine). xVM Server can be administered from any platform running a decent web browser.
The other difference between xVM and other Xen-based hypervisors is the base on which it's built. Citrix XenServer is built around CentOS which is used for the Dom0 (the administrative domain). Sun have built xVM around Solaris, so benefits from the FMA (Fault Management Architecture AKA self-healing), Crossbow (virtualised network stack), Dtrace and ZFS.
There is a lot of cool technology in xVM Server and it's certainly worth a look.
Nobody wants smaller MTUs, but with 1 and 10 gigabit ethernet, they sure as hell want larger ones.
.... you know the support is good.
They will stay with you across time zones as the Sun raises if you have a major problem.
IANAL but write like a drunk one.
Please enlighten us.
IANAL but write like a drunk one.
Still, that's not much of an excuse. Also worth noting that even if ZFS was GPL3 (Sun prefers GPL3 over 2, it seems), then that would still not be good enough for Linux. So yes, this is where Linus' choice of license is giving us some problems. Overall it was a good choice, but this is the bad part.
It was foolish and short-sighted for Linus to release Linux under the GPL v2 only, and not GPL v2 or later, as recommended by the Free Software Foundation. Now it is virtually (no pun intended) impossible to relicense the kernel under another license (the missing "or later" part), as there have been far too many contributors, some of whom are dead, in prison, or have otherwise vanished from the Community.
Sun prefers GPL v 3 as it does a better job of keeping the code free, particularly with respect to software patents, which, while not a problem for those of us lucky enough to be in Europe (not a problem for the moment, anyway), are certainly a concern in the US and other nations the US has bullied into adopting similar legislation.
As a result, technologies like ZFS are unlikely to ever make it into the Linux kernel. In the coming decades, as more and more technologies come along like this, Linus' inflexible licensing choice is likely to relegate the kernel to a historical footnote, where other kernels, licensed under either the "or later" clause (or other more permissive licenses) will continue. It's a pity, and I say that as one who has been using Linux since 1993, and will continue using it for the foreseeable future.
The Future of Human Evolution: Autonomy
I've benchmarked Windows and Linux under various hypervisors, and that is simply untrue.
Linux under Xen 3.3.0 gets 98% and better bare metal performance (both 32 and 64 bit paravirtualized). VMWare won't paravirtualize 64-bit (though they allude that someday they might), arguing it isn't necessary, BUT 64-bit Linux under VMWare gets nowhere near 98% bare metal performance.
Likewise, Windows (XP, 2003, 2008) under Windows Server 2008 with HyperV performs better than under VMWare ESX, and I say that as one who doesn't care for Microsoft products at all.
Xen and HyperV may have their issues (HyperV in particular doesn't support live migration, and won't until 2010), but performance isn't one of them. For someone wanting to Virtualize on the cheap, and maximize performance, Linux under Xen and Windows under HyperV appears to be the way to go...with Windows on VMWare if you need live migration today. Others may disagree or have other opinions based on their needs, but to religiously proclaim VMWare as the only solution and all other virtualization technologies inferior is, well, partisan, dogmatic, and contrary to the real-world experience and benchmarking of many of us.
The Future of Human Evolution: Autonomy
> only MTUs of 1500 bytes are supported (when did you see one smaller recently??)
We run 9000 MTU here on GigE which makes a difference for AoE & 9p
There are places where the networks are not touching,and there are places where they are-Boeing's Lori Gunter
I'm looking at setting up a new server in my house.
Currently I have a old AMD server running debian with VMWare server on it running XP which I can Remote into to do my Windows Only stuff. (Rather than waste space on my MacBookPro with VMWare). XP is doggishly slow (It's only 1.5gHz mobile processor).
I'm looking at turning that into an OpenFiler or FreeNAS machine (it has 2TB of HD on it) and getting a newer machine to help with iPod transcoding, other processing and virtualizing XP.
What should I be looking for in processors? Stuff that supports virtualization? Multiple cores?
Is this going to be a viable solution for me?
So... disclaimer. I'm a VMware employee, so I do know all about both these benchmarks (even if I had nothing to do with them). Agree the first VMware benchmark was quite skewed, looking at Xen instead of XenSource. The XenSource benchmark showed up, it showed Xen ahead in system-call microbenchmarks (hardware virtualization does well there, but lots of system calls with no I/O isn't representative of the real world) and more or less even on everything else. VMware approved XenSource's whitepaper for publication about two weeks later (which, BTW, is no longer on Citrix's website and not visible on Google). The comparison was not apples-to-apples - XenSource switched from Xen 3.0 to Xen 3.2 in the comparison, and didn't make any software-virtualization/hardware-virtualization tweaks. In other words, XenSource's benchmark was just as skewed as VMware's. And everybody who knows anything about benchmarking knows it.
The summary of that whole mess: XenSource / Simon Crosby got more PR mileage out of making a big deal of EULA restrictions than from any actual performance comparison. They never cared about a performance comparison - it was all a PR stunt to get a great big /REDACTED/ document posted to news sites / blogs.
VMware does not forbid negative benchmarks; they do forbid stupid benchmarks. Usually, some amateur runs Passmark 2D, which is a system-call microbenchmark that doesn't even keep time correctly in a virtual machine. Every single person complaining about that EULA has never bothered submitting results - almost all submissions get approved.
A witty [sig] proves nothing. --Voltaire
I don't know why Geordi LaForge has to have his vision system under GPL. Let the guy see with out all the red tape! You're giving him a headache with all of this paperwork!
The summary of that whole mess: XenSource / Simon Crosby got more PR mileage out of making a big deal of EULA restrictions than from any actual performance comparison.
Wouldn't it be in vmware's best interest to get rid of the idiotic EULA restrictions then? Trying to shut people up makes you look bad. Letting people publish stupid benchmarks and then demonstrating how they're stupid makes them look bad. Openness is always the answer. Sunshine is the best disinfectant. Censorship just makes you look like you have something to hide.
Give me Classic Slashdot or give me death!
...over at the 360is blog here. Once VMWare used to charge $5K per server, then Citrix halved that, before Microsoft came out with their offering for less than $40! Hypervisors are crashing in price faster than hard drive space.
AG.
Nice FUD.
No, this is not what happened at all. Simon Crosby (biggest blowhard ever), shot his mouth off proclaiming that VMware are a bunch of idiots, but he can't show it cause of the EULA. Well, unbeknownst to all his readers Xen had submitted their paper to VMware for approval, which they did approve and Xen published. It showed that Xen was competitive in most of the benchmarks, but fell short in a number and beat ESX in only 1, SPECjbb on Linux.
Good luck finding anything from this whole exchange, Citrix purged there blogs of the entire ordeal. Here is the paper WITH the data, no redactions. I am not seeing this "everywhere else Xen killed", could you point it out to me?
As a side note VMware is very liberal with their benchmark policy. As long as you actually benchmark in a sane manner they will let you publish no matter the result.
Q.
Or those who don't need support?
I bet if ken wanted virtualization, he would not pay for support.
OTOH, I don't think he'd use an off-the-shelf offering; he'd just take a weekend and code up a better one.
hah, you want to know what can't keep time correctly in a virtual machine, your fucking vmware workstation
Just curious. When's the last time anyone saw a (signed) post from a Sun VP on /.?
I've played with Xen, we use zones in Solaris, and I've used Microsoft's Virtual Server offering, but only VMware lets me do the one thing that no one else does: Put up a machine *fast*. I mean, from nothing to a fully working Linux/Windows/whatever machine whether it's a clone from an existing guest, or a brand new one.
I have a lot of projects that are ephemeral; we need a box to test something on and boom, we have a virtual machine that runs pretty darn fast and when the testing is done, we shut it down. No muss, no fuss. No other product on the market is so good about bringing up a machine, throwing additional "hardware" at it when necessary.
The other thing VMware rocks over everyone else is snapshots; I can create branches of branches of snapshots when my testing goes in all kinds of directions, and I can always roll back to any of them. I described it to a coworker as having the entire machine on top of a Subversion repository.
Don't take it as a flaimbait, but when VMWare itself will "keep time correctly" in the guest with respect to the host? No other VM/virtualization software I've used so far exhibits this strange "clock skew" behavior.
hosted on a vendor's own site...
No chance of bias at all...
XML is like violence. If it doesn't solve the problem, use more.
Except that managing a Solaris container is a lot more light weight than managing a VMware guest. It uses the same kernel version and kernel space between host and guest, and generally the container also shares /usr, etc. of the host via read-only loop back mounts. Services are separate, and so is /etc (and /var).
All you generally have to manage is the host OS, and the guest OS is more or less taken care of. Just look after the guest services, and not the guest OS.
Creating them is easy as well, as it just mostly copies the host (and re-creates /etc from a clean image). They work on SPARC and x86, and you can run Linux binaries in them through ABI-emulation. There are emulation modes for older versions of Solaris as well (8 and 9).
Sometimes you need a separate image with VMware, but sometimes you don't. Solaris 10 runs fine as a VMware guest, but if you don't want the overhead you have other options.
Use the tools that you need.
Sorry, made a small error, Simon's blog posts are still on Citrix's sight, but the paper's are gone.
Q.
A true bare-metal hypervisor would just run on bare-metal without any assistance of any OS, and would present the appearance of a bare-metal environment (usually one of exactly the same architecture, or architecture class, that it runs on) before any OS is even running on this. There would be no need for special OSes to be run on it. There would be no need for Solaris or Linux to be around. I could run only MS-DOS 1.0 on it, if it were for the x86 architecture.
now we need to go OSS in diesel cars
At this point in the game, virtualizing isn't a big deal. Its the suite of management tools aimed at the enterprise which separates the amateurs from the big boys.
Lets hope sun can step up to the plate and compete, and still stay in the OSS world in the process.
---- Booth was a patriot ----
I can create/destroy a new Xen VM in a matter of seconds, derived from some simple rules.
The management software in my case is a preview of an update to an IBM cluster management project, listing Xen support in the changelog:
http://xcat.wiki.sourceforge.net/xCAT+2.1+Changes
Creating 10 new VMs in my evaluation setup once configured was along the lines of
nodeadd v1-v10 groups=vm
rpower v1-v10 on
rinstall v1-v10
The VMs were off and installing my image within 30 seconds. It was kinda cool.
I love Sun, but this isn't viable yet. Vmotion? Clustering? A (relatively) easy to use GUI? Not even close.
Preaching to the choir :-). Not every American loves the antics of our President, not every employee loves the antics of their corporate overlords.
A witty [sig] proves nothing. --Voltaire
The summary of that whole mess: XenSource / Simon Crosby got more PR mileage out of making a big deal of EULA restrictions than from any actual performance comparison.
So why not remove the stupid EULA restriction?
Hatta already made the point, but it is worth repeating. This is really the same issue that Lessig's Change Congress movement is about - it doesn't matter whether cash contributions skews the political process or not, the mere existence of contributions is sufficient to cast doubt on the neutrality of the political process in the mind of the populace.
It does not matter whether the EULA restriction is used only to stop stupid benchmarks, the mere existence of the restriction creates the impression that vmware has something to hide.
If J.K.R wrote Windows: Puteulanus fenestra mortalis!
Good on you, Steve. Once VirtualBox fully supports USB2 and DirectX I'll move over in a flash.
No other VM/virtualization software I've used so far exhibits this strange "clock skew" behavior.
they all had it at some time or another. most are solved now, however
-Kz-
The FUD machine? Really? Fear may be justified, Uncertainty is reasonable if trends are visibly changing, Doubt is about observing bad trends. FUD is bad if unsupported, but is it unsupported? Lets talk cases.
First, why should there be so much potential FUD? It would be ever so simple to point out some third party benchmarks to easily make the point, rather than parroting the company line, but you failed to provide a link. I find plenty of references in articles to the EULA restrictions, articles that benchmark other VM platforms, but absolutely no other third party VMWare benchmarks. None. I'm sure there are some reputable comparative benchmarks out there somewhere since "allmost all submissions get approved" but I searched again today, still nothing. Why?
The only Fear I see is that somebody will test the products and publish results. Feel free though, you can download trials and test for yourself, good luck getting permission to publish. If you want to see the version from XenSource, google cache will provide. (Search for hypervisor_performance_comparison_1_0_5_with_esx-data.pdf) Here are some highlights:
As for Doubt, VMWare stock has been pretty much tanking. ("Tanking", like "killed" is a statement of opinion, not a technical term.) Really, I don't expect this to turn around for EMC's VMWare, and apparently neither do investors. Shares have gone from $125 in 2004, to $33.95 according to Forbes. Not what I'd call marketplace confidence. The Doubt part is certainly well established.
I like to plan to get good support in five years, but will EMC really be invested in supporting it if the stock becomes worthless? An independent company might reinvent itself, but a subsidiary of EMC, probably not so much. I suppose that is Uncertainty: will VMWare still be for sale and/or supported in 5 years? If I worked for VMWare, I'd be concerned about my job. Consider those already jumping ship: Diane Greene (co-founder, CEO), Mendel Rosenblum (co-founder, visionary and scientist), Richard Sarwal (exec from Oracle who went back to Oracle), and now Paul Chan, Vice President of Product Development.
Honestly, good benchmarks and solid technical comparisons by independent parties would go a long way. I say they aren't there because the Corporate Overlords fear them and aren't driven to improve the products, just sell them, and without improvements, VMWare will die. I expect EMC will abandon VMWare rather than investing in improving it. I expect that there are no current benchmarks published because EMC doesn't want them published. Please show me that I'm wrong. Please show me that EMC isn't killing what was once a great company with good products.
Show me the numbers, somewhere besides on VMWare's site.
B) Eliminate all the stupid users. This is frowned upon by society.
The separate guest image virtual disks are going to end up chewing up a lot of storage. I think a big issue will be keeping the libraries of these things organised, but it's certainly going to chew up a lot of disk, too. I think there is an argument for including data storage in the conversation -- lightweight is going to become important. De-duplicators are sort of like CVS but for much larger files and file collections, and live mostly in the SAN appliance space I believe. People like HDS and Data Domain play here.
Do not mock my vision of impractical footwear