Looking Forward to Hyper-V in Server 10

New year, new products! Some time in 2015, we’re all going to be graced with the newest edition of Windows and Windows Server, and along with them, Hyper-V. I wish I had a slick code name to give you, like “Viridian”, but it seems like most in-progress Microsoft products are now just code-named “vNext”. I’ve spent some time going over the published feature list. Some of the introductions will be very welcome. Some make me a bit less than enthusiastic.

Sole Sourcing

I’ve been burned more than once by writing about pre-release features and software conditions, even within a few months of release. The next version of Hyper-V is still quite a ways away, so there is still time for significant change. To that end, I’m only going to work with the officially published material on Hyper-V. Even that material could still be considered malleable at this point, but, in my opinion, it’s safer than relying on any third-party sources.

The Features I Like

We’ll start with those that I’m looking forward to.

Production Checkpoints

This is an interesting evolution of checkpoints (formerly snapshots). The first thing I want to point out is that they do not replace the existing checkpoint functionality; it’s just a different way of taking a checkpoint. There are two things of note here: first, and most important, Production Checkpoints are supported “for all production workloads.” “All“. They’re still not going to replace a proper backup solution, but now they’ll be the next best thing. The second important change is that they will, according to the wording of that document, tell Linux guests to “flush their file system buffers”. As far as I know, the current built-in and agentless systems don’t do that at all. This new power will give you a checkpoint that works like a backup that’s somewhere between crash-consistent and application-consistent. It won’t be application-consistent because not all applications’ buffers are tied to the file system buffer. It’s still about as good as you’re going to get on Linux at this phase. On this subject, I’ve also heard some rumors about changes to backup, but they’re not on this list, so…

Hyper-V Manager Improvements

Something we spend a lot of time on in the forums is explaining to people about why Hyper-V Manager doesn’t magically work when you run it as an administrator (because it is a computer-to-computer authentication, not user-to-computer) and how to set up constrained delegation or the Trusted Hosts list. Hyper-V Manager will now start working with alternate credentials, which means that at least some aspect of the computer-to-computer authentication is going away. That will certainly remove the need for constrained delegation. I assume that configuring Trusted Hosts will still be a necessity for those who insist on not joining their Hyper-V hosts to their domains, as WS-MAN does need some form of computer authentication. I certainly hope that this is how it will work, as I really hope Microsoft doesn’t start encouraging people to try to rely on workgroup-grade security. However, it doesn’t officially say so we won’t know until someone tries. At least for intra-domain authentication, this change will make things much easier.

Integration Services Delivered through Windows Update

This will be great, mainly for installation time. To have these updates synchronized with all the other updates will definitely be a blessing. Of course, this assumes that the Integrated Services are tested before being placed on Windows Update. The new feature might be a mixed blessing. For now, I’m cautiously optimistic.

Connected Standby

One of the features advertised for Hyper-V in its 2012 incarnation was that it would support sleep modes. Apparently, that wasn’t true. If it was, I didn’t encounter anyone who found the magic combination that would make it work. Why would I want sleep support on Hyper-V? Well, because not all instances of Hyper-V are production, that’s why. I use a “development” machine to work on CoreFig and all the scripts that I share here. It doesn’t get used even 2% of any given day, so why should it suck power 100% of the day? Sure, I can turn it off and on, but that’s pretty annoying when I just want to test one or two lines. I also had Hyper-V enabled on my laptop for a time, as I did quite a bit of development work there, as well. I’m the guy that likes to fold up the laptop and stick it in the bag when I’m ready to move. Enabling Client Hyper-V made that a bad plan for me, since I really didn’t want a portable heater. I also enabled Client Hyper-V on my desktop, which was a disaster. I don’t recall now if that’s because it would sleep and never wake up or if it wouldn’t sleep, but I do remember that I had to disable the role. What really makes it unforgivable for me is that sleep was a published feature of 8/2012 and it never really worked correctly.

Connected Standby is a new sleep mode introduced in Windows 8. It needs some pretty special hardware to function, so it’s still not going to work on any of the systems I have. However, many of you that carry around the latest and greatest laptops for demo purposes already have access to the features, so maybe you’ll get mileage out of it. For my personal purposes, I’d still like Hyper-V to allow my boxes to just plain old go to sleep because I don’t actually care if my test guests keep running, but this would be more welcome than the current option of going through a complete power cycle each time.

Rolling Cluster Upgrades

With Microsoft’s ever-shrinking release cycles and the community trust they’ve forfeited in the company’s ability to release quality software, an improved upgrade experience was an absolute necessity. What this feature does is basically let you run multiple versions of Windows/Hyper-V Server in the same cluster and freely move virtual machines between them. That will allow you to perform an in-place upgrade of a cluster without all the juggling tricks we had to do in the previous versions.

Multiple Virtual Machine Configuration Versions

This is essentially the technology that allows the rolling cluster upgrades to work. The same host can run virtual machines with the 2012 R2 format and the Server 10 format. This makes me a little happy, as you’ll see more clearly after you read the next session.

The Features I Don’t Care So Much About

With the exception of the first, I won’t characterize these features as “bad”, because that would be untrue. It’s just that I wouldn’t consider an in-place upgrade just for them.

Déjà-New Virtual Machine File Format

The more things change, the more they stay the same. First, we had VMC, and later VMCX, files that described virtual machines under Virtual PC. And then we got XML files in Hyper-V, which were awesome because we could understand what was going on without needing an interpreter and we could make all sorts of clever manual repairs when Hyper-V (or, more likely, an antivirus package) went into hack-and-slash mode on a VM. Well, apparently, Microsoft didn’t like that, because we’re going back to VMCX files in vNext. This is not awesome. Unless something is forthcoming or I missed a publication, the VMCX file format is closed. That means anything we discover about it will be through hackery. Manually fixing them or making minor adjustments will immediately be off the reservation and completely unsupported. This is not progress.

The official line for this regression to vLastLastLastLastLast is twofold. First, they claim that it is to “increase the efficiency of reading and writing virtual machine configuration data”. While I’m sure this is technically true, I’d like to know at what scale this efficiency becomes measurable, much less meaningful. I’m thinking there aren’t many installations that qualify.

The second given reason is, “It is also designed to reduce the potential for data corruption in the event of a storage failure”. This sounds more like the silly belief I’ve been fighting against for well over a decade, that files in a “binary” format are somehow inherently resistant to modification and corruption in a way that “text” files are not. It kills me how many otherwise intelligent IT people believe this nonsense. The difference between “text” files and “binary” files is pure semantics. A computer can no more read any given “binary” file without a dedicated parser than it can a “text” file because they’re both nothing but serially-stored bits whose organization is by 100% human-imposed meaning. The only “real” difference between the two is that “binary” files usually have a much wider range of acceptable unique bit-orderings than “text” files. When you think about that for a bit, it actually means that corruption should be harder to detect in “binary” files, not easier. I mean, if you “TYPE” a “text” file and your computer starts making beeping sounds, it’s a safe bet that the file is corrupted. For “binary” files, it’s perfectly normal. The notion that “binary” somehow provides superior stability over “text” files is like saying that Japanese is more stable than English on the sole basis that I can’t read or write Japanese without an external translator. The reason that the current XML reading/writing mechanism doesn’t have this ability “to reduce the potential for data corruption” is because there is no programming in place to make it that way, not because it couldn’t have been done.

At this point, I’m highly skeptical that this is a positive change.

Hot Add and Remove of Memory and Network Adapters

I think this is really about feature parity with the competition more than anything else. As far as I’m concerned, if you’re swapping memory and NICs in and out of your guests so much that it’s a show-stopper that it can’t currently be done online, then your provisioning skills are what need to be upgraded, not your hypervisor. That said, it’s a neat feature. I’ll certainly test it in my lab, but I suspect that I’ll use it somewhere around never in production. Oh, and as you’d probably expect, it’s only going to work for Generation 2 VMs.

Enhancements to Storage Quality of Service

I kind of feel bad about putting this feature in this section because it’s a good thing. It’s just a bit before its time. Unlike the existing storage QoS settings that can limit a VM in any situation, this new feature set requires that VMs live on a Scale-out File Server. That means Windows Server storage. Microsoft is certainly making inroads in storage and I expect adoption to continue to climb. However, I expect the saturation measurement of shops that are ripping out their functional EMC and NetApp deployments to replace them with SOFS is probably somewhere around zero percent. So, it’s good that Microsoft is getting out ahead of the game and providing features that will be desirable when people do start thinking about sending the big vendors and their forklifts away, but I don’t know that it’s going to make a big splash in 2015.

Honorable Mention: Linux Secure Boot

I have no real opinion as this feature just doesn’t affect me that much. The purpose of Secure Boot is to guard against corruption in the boot process of a guest, such as the changes made by a rootkit. The “firmware”, which in this case is Hyper-V, maintains a secured database of acceptable boot image signatures and won’t allow any boot image that doesn’t match the list to start. The problem is, there are a lot of legitimate operating systems out there that aren’t in Hyper-V’s database. This change extends this protection to some Linux guests. I’m not sure why this wasn’t available in vNow, but there it is.

A Lot Can Happen in a Few Months

As I said before, there’s still quite a bit of time between now and the actual release date. It will be interesting to see how solid these features are in comparison to what we’ll get in the final product.

I’ve refrained from speculating, but you’re certainly welcome to speculate all you want.

Your Thoughts

What do you think? What are you looking forward to most? Leave a comment and let us know!