Microsoft’s Windows Virtual Desktop service is now generally available | Hacker News Reader

Microsoft’s Windows Virtual Desktop service is now generally available

techcrunch.com

87 pointsindigodaddy63 comments

I came across FSLogix a few months before MS bought them and have been entirely impressed by how fantastic the product is. If you have ever dealt with roaming profiles and huge amount of junk in C:\Users\username folder, FSLogix is a blessing. In simplest terms, once installed on Windows (physical or VM, doesn’t matter), the entire user folder lives as a single VHD file on a network drive but appears to the local machine as a regular user folder. This means browser temp files do not need to be synced upon log off. In fact there is no syncing. Everything is accessible using just one network file stream. As someone who manages 200 users on Citrix, it is beautiful. Their Office plugin does the same with user’s Office profile, including Outlook. I haven’t played with the integrated FSLogix/MS offering but the pre-MS stuff is nothing short of a relief for sysadmins.

There's still large areas just 35 miles from DC that don't have high speed internet. The cloud is STILL just for people who never leave the range of high speed internet.

> people who never leave the range of high speed internet

Which is a vast majority of the US population. I agree that it is important to get broadband access for everyone, but asking companies to hold back services and features until that happens makes zero sense.

On top of that, if you want to drive companies to offer broadband everywhere -- having services people want to use everywhere that require broadband is pretty damn important.

It will have the reverse effect of drawing more people to where they can access those services and away from the areas where access to those services are limited due to inadequate internet services which will erode the motivation to provide broadband to those places.

I don't know about that. Here in Kazakhstan many villages gained relatively high-speed ADSL (up to 8 Mb/s, usually 4 Mb/s) in the past 4-5 years. Most of them are not exactly flourishing either: imagine something like Detroit (from what I've seen), but on a much smaller scale. At least it didn't stop them from streaming the trendy Hollywood crap everyone talks about.

One of the many reasons urbanisation is trending upwards.

http://www.unesco.org/education/tlsf/mods/theme_c/popups/mod...

First of all, offices virtually always do have high speed internet, which is the primary use case here.

But even if you're working from somewhere else, remote desktop doesn't need "high speed" anyways. It sends intercepted OS-level draw commands rather than bitmaps, so as long as you're not doing Photoshop it's closer to terminal bandwidth than YouTube bandwidth.

For the vast majority of business apps over remote desktop, whether your internet is high speed or not is a complete non-issue.

And the basic latency (e.g. for typing text or clicking a menu) is the same as terminal latency too.

RDP quit forwarding rendering commands some time ago,[1] it turns out to be faster to just send a compressed image. On very low bandwidth links effects are turned off and a limited color pallet is used.

You're still right -- the performance is good, and using normal UI controls you will typically not notice even a very poor connection. But this is because UI changes are an "eventually consistent" deal, and although the Windows terminal services implementation does send bitmaps it optimizes drawing for sending the resulting bitmap. Contrast to scrolling text on a terminal that may not update in a manner that allows you to react to it.

[1]: Somewhere on the FreeRDP mailing list. Sorry :(

Not to mention the fact that just about no program uses the old Win32 rendering commands anymore for anything relevant. They are just far too limited.

> And the basic latency (e.g. for typing text or clicking a menu) is the same as terminal latency too.

If "terminal latency" is the high watermark, it doesn't bring pleasant thoughts to mind. I find typing over ssh to be irritating. Not awful, but a clear downgrade from a local terminal.

Shit, my company is in a 8 million+ people suburb and we pay $5k/month just for guaranteed 50 mbps internet. And we are saturating it too easily as is :/ A gigabit connection was quoted at ~ $80k/month. (The line is already fiber, it's literally just throttling). Which is insane considering that's 12 peoples salary in a year.

Wow, where do you live?

There are large areas 35 miles from DC that have no public sewer or water either. (My parents’ house, for example, has neither, but does have fiber.) But what percentage of the population does that account for? Around DC, very little.

In Akamai’s 2017 report, Maryland’s %-age of connections above 10 mbps (easily enough to do Remote Desktop) was 79%, just below South Korea and above Switzerland. Virginia and the district itself were around 75%, the same as Switzerland.

And, more to the point, some people not having public water and sewerage systems doesn't stop us selling toilets or taps.

Well that analogy doesn’t hold because you still use toilets and taps in those well and septic areas near DC the author is referring to.

You can still use cloud products in areas where there's not fibre/ultra-broadband internet. You just have to put up with a little bit of shit every now and then - much like having a backyard septic!

For what it's worth I used to use remote desktop from across the world on dialup, and it was certainly usable.

I use RDP pretty heavily in low-bandwidth and high-latency scenarios. In the last couple years I've seen a lot more software and websites using "fade in/out" effects that are utterly excruciating. Aside from those the performance is great.

In many cases we've found that RDPing to a system across the country to browse the web was a significantly smoother experience than trying to browse the web locally on that same connection. RDP is astoundingly good at what it does.

This is for enterprises that typically use Citrix as the modern version of X Windows.

uh, where? The middle of the Chesapeake bay?

The area around DC, outside of the suburbs, is quite rural actually.

That’s pretty darn far. Are you trying to say Chantilly? Leesburg? Or more like Front Royal?

Anything beyond Manassas to the West and Quantico to the South, Leesburg, Warrenton, Culpeper, Fredericksburg, Spotsylvania, parts of MD to the north past Germantown, etc.

There's plenty of rural on septic systems and well water just an hour's drive out of DC.

umm. not exactly for that audience or people like you and me to be honest. i'd rather gmail at home than office 365...

this is for companies. with enough employees who are building on the microsoft stack and the office suite. god knows we would rather they deal with their updates themselves. and besides, way things are going virtualization is being cheaper than buying hardware...

I live within 15 miles of the main Microsoft campus in Redmond and only have access to shitty DSL. Microsoft has always taken the bet that everyone in the future will have access to high speed internet, but without ever raising a finger to make sure that future actually happens.

Shitty DSL is fine for remote desktop.

That depends on your definition of shitty ADSL. My old ADSL wasn't just slow, it was unstable and would regularly starting lagging out / having massive pings which would then go away and come back. Three years of talking to the ISP and I got nowhere. I now live in an NBN area and get Fibre to the Curb with 80MBPS down, I can now truly work from home.

I can see the security benefit of a virtualized desktop, and in some industries I understand that that's enough to justify this kind of solution.

Aside from that, I don't see so many benefits. The company still needs to provide to each employee a device, and training. I don't see many savings on this side. Companies going for cheap devices will still go for cheap devices, while those looking for higher end devices won't settle for cheap alternatives. Besides, SSD and CPU are now very cheap, so you can get very good configurations for cheap. On the other hand, batteries are still expensive, and virtualization might drain them more than using the local CPU depending on the use case.

If the added value is the easier configuration phase and maintenance for IT (software side, as devices still need maintenance), then ChromeOS seems like a much better approach. Instead of delegating to a third party the whole OS (virtualized), the customer only transfers the management of the services. And to be fair, if security here is the main concern, then ChromeOS seems still better positioned, both because data is on the cloud, and because of how the OS is "restricted".

At the end of the day, WVD seems like a way to sell (also) azure services on top of windows. So I wouldn't be surprised if companies evaluating this option will see an increment in costs.

At this point I wonder if it's really worth, especially considering that if for any reason the company cannot keep paying Microsoft for the service, or it's cut off from the service (e.g. non us company -> sanctions), basically the whole company is back to paper in a matter of minutes, while with non-virtualized solutions there's a much bigger safety net.

As someone who spends significant amounts of time on the road, a setup like this can be a godsend.

You mention more CPU, more SSD, more memory, just keeps getting cheaper and you’re right. You can get a pretty amazing laptop for not much more than an iPad Pro and it will handle nearly any workload. The problem is now you have to carry that and all of the batteries and accessories around. It’s so much easier for me to bring my iPad Pro 11” with a keyboard case and remote into a desktop environment for those workflows that either require a high end machine or Windows. I imagine a Surface would be even better.

Personally, I just don’t want to carry a massive laptop to get compute if I can access it remotely just fine. So that’s what I do. I have multiple VMs running Windows or Ubuntu on a high end Ryzen machine with multiple Nvidia GPUs. That’s significant more power than I could ever hope to bring with me on the road. For me the only major setback is screen real estate. Hoping one day some of these HMDs can fix that.

If Stadia and xCloud really do solve the latency issues inherent in client/server computing - and show that it's economical - one could almost imagine Microsoft launching a consumer version of this with a "terminal" that was little more than a screen, keyboard and network connection... Like Sun Microsystems reborn.

The whole concept would, of course, enrage certain sections of the populace including most of HN's readership, but I can see it happening.

How do you beat the speed of light? That's ultimately what we're up against here. High-frequency traders lay down special fiber lines between the exchanges to get the absolute fastest possible ping times, at enormous expense[1]. That gets a 13 ms ping time from NYC to Chicago.

My experience with a Steam Link, hardwired on a local ethernet makes me skeptical that any of these cloud gaming platforms will really be practical.

[1] https://www.telegraph.co.uk/finance/newsbysector/banksandfin...

13 ms is not noticeable and almost everyone is much closer to Google or Microsoft machines than NYC is to Chicago. The bigger but more tractable issue are the steps between your machine and your ISP.

A lot of traders are actually switching to wireless links since RF travels faster than light in a fiber. But some fun stuff happens then https://www.bloomberg.com/news/articles/2018-07-03/this-heat...

It travels faster because it travels "as the crow flies" instead of snaking through all the conduit for a fiber connection.

If nothing else, those dedicated services are going to have specialized video encoding that is very, very fast, and possibly even integrated into the game render pipeline. Maybe Steam Link is better on a really beefy PC, but I tried the remote play built into my PS4, and it was pretty bad, even over wired LAN.

I have to imagine at that point that at least some of the latency is the cost of encoding the video.

Even pretty hard core gamers seem to think that 40ms or less is good for gaming. So that leaves some leeway to work with I guess.

Tried steam link on local eth too with same conclusion, but later realised that was probably cause I had x264 not 265 which apparently makes a massive diff

Stadia isn't for shooters or esports though, it's for Age of Empires and Civ.

You know, I’d pay for that service. For example like $4.99/month if they give me say 200 hours per month. (I haven’t done the math or the research if that is economically sane).

Some idea.

I must be missing something. It's not immediately clear to me, even after reading the article, how this is different from just procuring a Windows OS VM in Azure. Pricing?

Was wondering the same thing. Is this Azures version of AWS Workspaces?

And the cycle restarts as we head back towards thin clients...

I have a friend who works at a big Corporation (100k employees) where everyone has a thin dumb client and windows desktop running on a server. It works great, except video stutters if not small sized. Way of the future? Its an efficient way to pool resources, and prevents people physically stealing data.

Been running a setup like that for over 6 years. It’s great. Support is ridiculously easy. Desktop refreshes aren’t bad either and that’s one of the most difficult projects for an IT department to pull off. We’ve done a couple of them now - once to add RAM, disk and CPU and another to change the desktop image. Deployments and big jobs like opening new offices or big moves are simple.

The video is easy to fix with a fast network or Apex PCOIP accelerator cards, although those are being replaced by expensive server-side GPU’s these days. We have some google earth users and a few others with gpu tasks. A couple Apex cards in a couple of the servers took care of that.

The only thing that completely sucks is USB storage devices. They are painfully slow. Most people accept it because they are rarely used in our environment and everything else is faster than any other system they’ve had.

Printing with terminal services still kind of sucks and for large print jobs there's still third party print services, but otherwise I'd agree.

It’s not terminal services. We are using VMware horizon. Printing is fine. USB scanning is a little slow but most people don’t seem to notice.

I worked for a while at a company which did this and, while it worked, everything was slightly worse (for me as a programmer -- I'm sure support was much easier). Noticeable latency while doing normal things like scrolling through code or switching windows. Every so often there would be a latency spike which, because it happened infrequently but unpredictably, was surprisingly disconcerting. I ended up having to retrain myself to use "low bandwidth" actions, like pgup/pgdn rather than using a scrollwheel. I don't think it significantly impacted my performance, but I often found myself wishing I had a local desktop.

> Way of the future?

Why not? Everything old becomes new again at some point.

Like X Window

The first time I ran a program on a remote computer, and used its UI over the Internet, was 1992. In fact that’s how I discovered Mosaic. Running random executables in the bin directory of lab machines.

Has problems when your staff spend a lot of time mobile though. Our staff have laptops and are quite often travelling on trains, aeroplanes, in offices and hotels where their guest WiFi is borked. While the laptops have a 4G chip, and connectivity is improving year by year, it is still not feasible to rely on remote desktop only for our use case. Local desktop and offline file sync is still the only way to go here.

Actually spirit of the past, aka timeshare mainframes.

I pretty much live inside Remote Desktop, and the only fault I can find with this is that it starts at 100 users (in the Azure pricing calculator). Having something like this for SoHo (1-5 users) would be a killer app in my neck of the woods.

So... this is Terminal Services as a Service?

I wonder how it compares to AWS Systems Manager Session Manager.

How does app installation work? ie. If I have a license for some software, can I rent a virtual desktop to run it on?

Is this Azures version of AWS Workspaces?

Did anyone even test the process. It has failed a dozen times in different way.

“Generally “ available? So does that mean it’s available sometimes, but not to everyone or not always?

"General" availability means available to the general public, basically. as opposed to beta/insiders etc.

Thanks, was wondering about that

Reply on news.ycombinator.com