The next feature update for Windows 10 (known in the Windows Insider Program as 19H2) will be a scoped set of features for select performance improvements, enterprise features and quality enhancements. To deliver these updates in a less disruptive fashion, we will deliver this feature update in a new way, using servicing technology (like the monthly update process) for customers running the May 2019 Update who choose to update to the new release. In other words, anyone running the May 2019 Update and updating to the new release will have a far faster update experience because the update will install like a monthly update.
This service pack-like release is scheduled for September. I do have to say though that I am starting to miss the forest through the trees when it comes to Windows and its updates. I understand why things have to be so complicated – Windows is used in many different environments, and each environment requires unique updating rules – but it hasn’t exactly made things easier to grasp for consumers.
This is truly great. The monthly update process is extremly quick nowadays and causes about a minute of downtime on my 3 year old machine. The half-yearly upgrades used to be 45 minutes of downtime but became 20 minutes and will now be 1 minute. Of course the actual installation requires more time, but the actual downtime (the part regular users notice) has basically become 0.
And there isn’t anything complicated to grasp for consumers. When Microsoft thinks their machines would be best served with the newer features they will be offered this upgrade, it installs quietly and eventually reboots (now quickly) just like it does every month. It is only “us geeks” that actually keep track of the version of Windows that runs on their machines (and “seek” out new upgrades) that wonder why they haven’t received upgradeX.
I am wondering how they made such a massive improvement though. In the past updates (small .msp patch) were installed very differently from upgrades (massive .esd system-image). If this happens to be a side-effect of that “7 GB* reserved upgrade/installation partition” that was introduced in 19H1 that would be well worth 7 GB of space to me!
* https://techcommunity.microsoft.com/t5/Storage-at-Microsoft/Windows-10-and-reserved-storage/ba-p/428327
If only it were as nice as you describe. First, the new update process eats CPU like there was no tomorrow, and it is an issue on low end hardware (dual core Pentium) . Then, the update messages are damn useless. Those mysterious messages explain nothing, but if you do not reboot immediately when a reboot is required, you will enter a world of pain, my friend. You can get random failures of applications with no real explanation. I lost quite some time trying to figure out why my Firefox, Outlook, and many other, were failing for no good reason; I had applied an updated, delayed the reboot for a couple of hours, then totally forgot about it.
The new update process indeed uses more CPU because it uses smaller delta-patches (less IO, more CPU). Of course these updates are executed on low priority so they shouldn’t cause any slow down for actual work. My impression was that:
* a full offline update (dism) only took about 2 minutes (and because it is basically a single-task at that moment it takes up all the CPU and IO that it can during the downtime) followed by about 30 seconds of online downtime (30% to 100%).
* The old updates took about 5 minutes of heavy load (IO) and longer reboots.
* The newer updates take about 10-20 minutes with moderate load (IO) and shorter reboots.
Basically the newer updates take longer but you can keep working while they apply in the background. Roughly the same applies for upgrades but I have only installed a few insider previews this way
* The old upgrades took about 45 minutes with half of that in downtime, The newer upgrades take about 2 hours but with only about 5 minutes of downtime.
And the random failures really shouldn’t happen. Updates and upgrades are applied with atomic transactions on the filesystem. Until the system reboots you should be “getting the old versions”. If this doesn’t happen you might have disabled system restore points and file history
We have been reading this for years, that this will be the breakthrough that makes windows updates fast and reliable. The sad truth is: The windows update experience is a nightmare and this will change nothing. May be updates will fail a bit faster. This is especially true of you manage windows updates in an enterprise environment with a broader range of hardware/software. At least 20% fail with cryptic error codes you have to google. Why? What is so hard to give me a human readable error message? On my linux servers I issue two commands rarely have to wait more then 3 minutes to download and install and failures are rare. (scnr).
Microsoft has no quality control, a beyond broken update architecture, and no clue how to fix it.
At least 20% fail? You think that in an enterprise environment there is a broader range of hardware?
If at least 20% would fail, hardly any home user or business user would have been able to upgrade from 1507 to 1803, yet hardly anyone is on 1507 and almost everyone is on 1803 or newer
In Enterprise there are almost always a bunch of hardware-configurations that are distributed at any amount of time depending on the desired functionality. Enterprise tries to keep this number as low as possible because every variation cause support/maintenance/testing/etc costs to go up. Basically they either buy Dell/HP/Lenovo/Apple and not much else. Meanwhile homeusers built their own pc’s, buy Asus/Acer/Medion/Target/Whatever-is-on-sale in hundreds of different variations. Or more simply put: In enterprise the one sitting left and right of me will use the same hardware while at home my neighbours will use different hardware.
Windows 7 had a broken upgrade architecture. No more Service Packs were released and you had to install literally over 100 updates even if you were on the latest Service Pack. Recently they retro-actively fixed this by supplying cumulative updates similar to the CU’s from Windows 10 and now the failure rate on our WSUS/SCCM environment is back down to about 2 percent (mostly people putting their laptops in standby during the update). On the Windows 10 side you can clearly see the difference between the retracted 1809 (we never deployed that one) and 1903 (we just started deploying this and sofar it is just as succesful as 1803 but installs with less downtime)
Windows Updates aren’t as effortless as on Linux and Server Updates in general are far less problematic compared to Client Updates. However Windows Updates AND Upgrades are getting better and better lately with less problems and less downtime