As VMware steps into the virtualization limelight in San Francisco at its VMworld conference Sept. 11, Microsoft is trying to steal some of that thunder by making a number of announcements of its own. Microsoft is announcing that it plans to deliver the first release candidate of Windows Server 2008 later this month. The first Community Technology Preview of Windows Server Virtualization, code-named Viridian, will be released at the same time, Larry Orecklin, Microsoft’s general manager of marketing for System Center, said at a media event Sept. 10.
Will WS 2008 come out just in time to steal the thunder of Samba 4? Samba 4 will enable users to setup an Active Directory commpatible server – ie., operate in a WS 2003 environment (including being the Primary Domain Controller). I assume WS 2008 will introduce just enough incompatibility to SMB and AD to cause Samba 4 users to again be behind the times with Microsoft’s current server line. On the other hand, maybe the Samba folks are already following this and will be able to fulfill the role of a 2008 server. Just wondering!
I really think that Server 2008 is going to work better with Open Source enterprise apps than 2003 actually, if I am permited to write about it I will try that right after RC1 is released.
(Depends on the NDA)
The easier way to avoid the problem is move to something like iFolder or NFSv4 to avoid the version/revision compatibility issues.
iFolders, NSFv4 and samba are not interchangeable products. They solve slightly different problems, and have different strengths weaknesses and features.
How is it different? smb and nfs are interchangeable. Use OpenLDAP, NFS and services for Unix on the Windows client, and voila, problems gone.
Well, historically Samba has been about a generation behind Microsoft due to the fact that they are essentially reverse engineering the MS SMB/CIFS protocol using network packet analysis. (They’ve discovered some really cool things about Windows as a result too. Just check their docs and you’ll see some of the treasures.)
So, unless (as the other poster suggests) Microsoft plays more nicely this time around and doesn’t change their SMB/CIFS protocol, then it will likely happen again.
But that’s not really a big deal, as Samba would still be able to interop with the new protocol much the same way that WS2003, Win2k, WinXP, and others will. So they won’t be left out in the dark, just not able to run as a PDC/BDC in a WS2K8 network; but it is not likely that there will be many Win2k8 networks for a while as businesses will try it out in test labs, etc. before putting into production.
So, it’s all up to how much Microsoft modified the protocol; on the other hand, given that Vista has been out for a while now, it is possible that they could have reverse engineered some of the Vista-to-Vista specifics and have a head start – who knows, they may have even gotten a Win2k8 beta too (not likely though).
Best head over to the Samba mailing list for an official and real answer.
There’s no such thing as a PDC in an AD environment.
That doesn’t mean it’ll be trivial for the samba team to add win2008 compatibility, but hopefully the main change will be extending the LDAP schema.
although there is no PDC in the traditional sense, there is always the primary/first server. This server is a little bit different to other servers on the AD.
As for windows 2008, windows 2003 was an excellent release. Very fast rock solid, ive only tried out the win2008 CTP for a little bit, but first impressions are good. It’s still pretty amazing that the company behind Windows Vista is also the company behind Windows 2008, as so far i don’t really think much of vista still.
I suppose it comes down to that Microsoft is a good company with lots of very bright people that only seemed to be used when Microsoft wants to fend off opposition. I believe that the windows after Vista R2 will be incredibly good due to the mounting competition from Mac OSX and Linux.
However back on track, Core services look good for data centres, lowering the attack surface. I’ll be interested to see what sort of performance we will get from SQL Server 2008 and Win2008. As i think Win2003 and SQL 2005 are one of the best databases around, very quick/reliable with some great tools.
Although it would be nice to have a 2008 release for the three products at the same time (The last one i can remember was 2005 (VS, SQL Server and Biz Talk). I hope Microsoft doesn’t sacrifice Win2008 quaility for marketing.
I have to give MS some credit they are actually trying to clean up their product line.
Windows Server 2003 was head and tails better than W2K Server, I see Server 2008 to win over some other converts because in all reality in the DataCenter it is Windows, RedHat, or some Unix variant AIX, HP-UX, Sun Solaris.
is one of the coolest features of Windows server. It is a true lightweight hypervisor based virtualization for the most heavyweight server needs. Well ok may not the most heavyweight servers in the first release but it’s ability to provide true SMP support to virtual machines, a new virtual bus based device IO model seems very cool and I think would provide better performance than competing vmware and xen.
is one of the coolest features of Windows server. It is a true lightweight hypervisor based virtualization for the most heavyweight server needs. Well ok may not the most heavyweight servers in the first release but it’s ability to provide true SMP support to virtual machines, a new virtual bus based device IO model seems very cool and I think would provide better performance than competing vmware and xen.
Why was this post modded down to -1??
Seriously some people need to get over themselves on this forum.
It wasn’t modded. Highly trusted users start at +2, regular users start at +1, untrusted users start at 0, and those who are most often modded down start at -1. When a user leaves a good comment, mod them UP and help them repair their trust.
It wasn’t modded. Highly trusted users start at +2, regular users start at +1, untrusted users start at 0, and those who are most often modded down start at -1. When a user leaves a good comment, mod them UP and help them repair their trust.
Thanks for the explanation!
is one of the coolest features of Windows server. It is a true lightweight hypervisor based virtualization for the most heavyweight server needs.
One of the biggest problems that you have with virtualisation is the hardware and the OS you run your hypervisor on as the single point of failure, and you find that out when you do it. I fail to see why I should get excited about running a hypervisor and virtual machines on top of Windows, and all its inherent patches, built-in software and instability, when I can run ESX (or even a stripped Linux distro and VMware Server) and have far, far fewer problems.
Well ok may not the most heavyweight servers in the first release but it’s ability to provide true SMP support to virtual machines, a new virtual bus based device IO model seems very cool and I think would provide better performance than competing vmware and xen.
I very much doubt it. Running a hypervisor and a bunch of virtual machines on top of Windows is just not a good idea. The whole idea is that the OS or platform that you run your hypervisor on has practically nothing running on it, not just for performance but for stability reasons. Microsoft just can’t offer that. I can also get SMP support in ESX and in VMware Server now. I’m also not going to get excited about running Linux guests on a Windows server as a result.
I don’t really consider Viridian even to be much of a competitor to VMware Server, especially considering that it’s infinitely far less proven. I haven’t looked at the details, but I’d be very surprised if there weren’t some licensing shenanigans going on in there as well.
In the Linux world, Xen looks to be on the road to nowhere because it’s getting too far up its own backside as to the technical solutions it comes up with. KVM is far, far more straightforward, and I bet companies like Red Hat are just desperate to see the back of Xen and the support they have to provide for it. This will take some time to sort out before people really start using it in great numbers.
What are you basing your conclusion that
from the above statement on?
http://blogs.technet.com/aralves/archive/2007/02/28/longhorn-hyperv…
What are you basing your conclusion that Microsoft just can’t offer that. from the above statement on?
Microsoft appear to be going the same misguided way with a privileged guest, as in Xen’s privileged dom0 domain, although VMware seem to be heading in a another broken direction:
http://en.wikipedia.org/wiki/Windows_Server_Virtualization
A hypervisor instance has to have at least one Root Partition, running Windows Server 2008. The virtualization stack runs in the root partition and has direct access to the hardware devices. The root partition then creates the Child Partitions which hosts the guests OSs.
As has been pointed out in other virtualisation articles here, this approach is just loony even just from a stability point of view. Of course, that privileged guest can only be Windows, so there you are.
Roll on KVM, that’s what I say. Then we might get something sensible.
Sounds like you have completely misunderstood the approach being taken; the root partition is +not+ running in the guest OS.
Sounds like you have completely misunderstood the approach being taken; the root partition is +not+ running in the guest OS.
Nope, I haven’t misunderstood.
What are you basing your conclusion that
from the above statement on?
http://blogs.technet.com/aralves/archive/2007/02/28/longhorn-hyperv…
IIS 7. Period.
This version of IIS 7 will finish off the asp.net -> IIS integration story…a lot of folks don’t realize that IIS v < 7 is actually extremely customizable via ISAPI, but that entails digging into nasty C++ ATL code. Plugging something into the IIS 7 pipeline will be as easy as writing an httpModule is now in asp.net.
Windows 2003 + IIS 6 is something MS got pretty close to right on the first release, so expectations are pretty high for 7, as they should be. We’ve been doing some internal testing with Win2k8 betas and have seen some pretty substantial performance gains…and these are just betas. We expect the RTM to be an absolute screamer…think less servers in your web farm (hint, less TCO).
IIS 7 is surely an hot piece of software. We’re excited too to put our hands on it and, given the good tracks of IIS6, it will be for sure something we will be happy about.
We’ve been doing some internal testing with Win2k8 betas and have seen some pretty substantial performance gains…and these are just betas. We expect the RTM to be an absolute screamer…think less servers in your web farm (hint, less TCO).
Forgive a slight piece of skepticism here, but we hear this at every single release of Windows Server. Do more with less, less TCO etc. etc. The usual buzz-phrases.
“Forgive a slight piece of skepticism here, but we hear this at every single release of Windows Server.”
You’re forgiven. IT by definition is always extremely skeptical (at least they should be, ideally), which is why we test, benchmark, try to break, etc every new piece of software…especially when it’s something as critical as an application server which will end up supporting a good portion of our infrastructure. Why would we dole out hundreds of thousands of dollars for new software which is only marginally better than a tried and true solution (in this case, Windows 2003)?
In the case of a web server, even if you only see an increase in throughput of a few milliseconds per request, that adds up to huge gains over the course of a workday. We’re seeing increases in both throughput (# of requests handled/second) and higher cache hit ratios. The only spec we’ve noticed any non-trivial increase in is memory requirements, but that’s to be expected and is a small price to pay (memory is cheap).
Our TCO in the short term scheme of things will be lower due to less licenses needed…we’re probably going to be able to remove 15% of the servers in our web farm, and that’s a conservative estimate; we’ll have better numbers after the RC since a lot of debugging code will be stripped out. I’d say less licenses needed is a rock solid indication of lower TCO.
Microwho?
Among many other things, one of the major changes for 2008 is vastly improved CIFS performance in very high-latency WAN connections by heavily optimizing the protocol – they’re not just changing it to make Samba incompatible. The enterprise WAN accelerators (Cisco WAAS, Riverbed to name two of the more popular) deal with this issue currently using various schemes. The accelerators work and do all kinds of other cool stuff but, for many people, they are purchased specifically to optimize CIFS. The cost savings alone there can be significant.
Another think MS is doing is actually making the OS LIGHTER than before. 2003 already was like that compared to 2000 in my experience. 2008 has a console-only mode, for instance, that’s ridiculously lightweight. It’s like the enterprise Microsoft OSes follow the inverse trend of their home OSes bloat-wise (Vista was ridiculous).
Vista with SP1 will actually gain the 2008 kernel (or most of it) – SP1 will be a huge change for Vista. I don’t know if they’ll put out a patch for the older desktop OSes so they can use the new CIFS code, I hope they do. Of course, doing so for 2003 server might prove counter-productive sales-wise…
About whether Viridian will be able to compete with ESX day 1 is another story altogether. Remember, ESX is the enterprise VMware product. I’m almost certain most people that have left comments haven’t actually deployed ESX but instead most have played with Workstation and Server (I deal with ESX professionally, among other things…)
ESX is nowhere NEAR VMware server. The performance difference, especially under load (as in many, many VMs in a single box, all doing useful work), is staggering. Not to mention VMotion, DRS and HA. The management is amazingly easy. You can script everything. It’s practically bulletproof. It’s deployed in mission-critical and sensitive environments such as banks, hospitals, airlines, hosting providers, the list goes on. And with EMC owning VMware, it’s not like they are a mom-and-pop shop. The product was good before EMC bought it and keeps getting better. I have the list of enhancements for 3.5 and extremely cool stuff is coming out.
There’s a reason VMware can charge for it…
Then again, I’ll be one of the first to give Viridian a spin. If anything, I love testing stuff, and, like 2003, it might prove to be a pleasant surprise. Maybe if they run it on the console-only versions of 2008 it will work better.
D