Microsoft's corporate veep for enterprise puts the boot into boot times We formally bumped into Brad Anderson, Microsoft's corporate veep of Enterprise Experiences, at the software giant's London offices recently amid excited spurtings about the Windows Virtual Desktop (WVD) and the newly de-Windowed Microsoft Defender. Andersen is both a Novell refugee and, as he approaches his 16th full year at …
> Anderson used the example of an unnamed large US bank: "They had more than 20 agents on their device, they went down to just two. And their boot times went from three minutes to 20 seconds."
Please, please, please name these two agents. My employer has a ton of shit, but mostly S******* Endpoint Protection running and the first 3 minutes after boot are spent at 99% CPU while it "endpoint protects" me.
We just have Kaspersky on our machines. My ThinkPad takes around 20 seconds to boot and another 15 to log on.
At home, the BIOS screen on my Ryzen desktop takes 3 times as long as the Windows boot process (~10 seconds). Logging on takes another 3 - 4 seconds. My Spectre X360 (Skylake) takes around 20 seconds to book and 10 seconds to log on.
But I don't fill my PCs up with crapware.
OK, I'll say it - my suspend and reboot actually works.
The majority of the delay for me to get going is the time it takes to enter my password.
I rarely need to reboot - I can go a week or two without, and even then only do it because it's good to clean out the occasional process.
It ought to be possible to get Windows up to that level.
Agreed. Once the things are actually booted, they stay on, going into suspend mode when not in use, until the next patch requires a reboot.
I guess the PCs get rebooted once or twice a month. The rest of the time, it is suspend mode, which has them back up and running in a second or two.
Tell you what, I'll watch your corny adcasts if you promise to wear it throughout each show.
"With Windows 10 you get faster boot times BEEEEP reliable updates BEEEEEEP and increased productivity BEEEEEP. and your users will just love al the new features of Office 365 BEEEEP Your system admins can provide guaranteed uptime by combining the power of Active Directory and Azure BEEEEEEEEEEEEEEEEEP"
‘the unfortunate exec "had been late to a Skype for Business call because it took him eight minutes for his PC to boot" .. perhaps the finger of fault gets directed at a Redmond-based software giant. Anderson conceded: "And they could be blaming Microsoft." .. The solution, he said, is to (unsurprisingly) go all-in on Microsoft's tech.’
Our stuff® sucks™ and the solution is to buy more of our stuff®
It's sandwiched between the insanity of the UEFI/BIOS and the glacial idiocy caused by the 3rd party bloatware that is hot on the OS's heels.
Honestly the hardware end infuriates me more, as I can disable or remove alot of the bloatware, most of those boot agents aren't doing anything I need, and many are doing things I need them to NOT do. +1 to the statement above about SAV, it's cancer, and Windows Defender runs circles around it.
Most motherboard makers seem to labor in the illusion that I need to be able to still install a copy of TRS DOS and a parallel SCSI 2 cdrom drive on my M.2 socket 2066 Xeon motherboard, and insist on wasting an eternity loading legacy crap that will never again be seen in a modern computer. Only a few manufacturers let you knock that nonsense out in the BIOS and get truly optimized fast boot option.
The article was talking about desktops too, the server space is even worse. On the very bad days I still have to reboot a domain controller during daylight hours the windows load time isn't what I'm sweating, its the intolerable POST process from the hardware. I get that in some environments you want the redundant hardware scans, memory checks, and device based BIOS agents to load every.single.time, but in other cases, you need the hardware to come up as fast as possible. I'd pay and extra 500$ a box and switch server vendors just for a BIOS that POSTs about as fast as my desktops do.
Ditto for my switches, I have a stack of cisco 3850s that I swear take 15min to restart. They also are too cheap to include a power button, so if the firmware is buggy(it is) you gotta yank the power cables to unfreeze them. Very professional and high end, like those home cable modems.
This represents the largest probable threat in out disaster plans, even in the event of a simple power drop that out runs the UPS(or we shut things down to protect the equipment) we have to wait for switches, then load the storage arrays, then servers and the Call Manager. That chain is depressingly 45+ min from a cold shout down to working phones and internet. You can't even slip that in on your bosses lunch break :)
Sadly, my smaller clients that were running on non-enterprise hardware can get their dumb switches and glorified gaming desktop server tower up and serving asterisk in 8min flat. Sadly enterprise gear seems completely oblivious, and only willing to push redundant hardware at four times the price and four times the startup delay. Seems like a missed sales opportunity.
Or is their a fast boot server/switch vendor I don't know about?
Such slow booting switches cause some really huge problems. There are many places where the actual devices are up and running, waiting for network for ten to twenty minutes before the Cisco switches deign to do the link negotiation properly.
By which time the end device has concluded that the network switch is very broken and dropped down to 10 base T half duplex, because nothing else worked on the fiftieth try.
And the slow switches all seem to stack on each other.
A comes up, waits for B to be in a certain state before handshake can begin. B comes up but requires C to be available before handshake can begin. C is busy making sure it's able to talk to some DB server, a firewall, and the cheese vending machine. Meanwhile, the mainframe has been sitting there all the time telling everyone that these new systems are shit, dontcha know.
Having spent years involved in getting fault-tolerance, optimised start-up to fully active times for mainframe systems and the applications on them, to minimise the time it takes from out to back in business, how did we end up going from my 8 minutes to fully active to 45 plus minutes to layer everything back up? How is this an improvement? Who is dumb enough to think that enough redundancy means no outages?
I'd pay and extra 500$ a box and switch server vendors just for a BIOS that POSTs about as fast as my desktops do.
The problem with servers is they usually have RAID BIOSs and so on to initialise and boot, which take a while. But I'm not sure I'd be happy with a RAID controller that didn't check for volume consistency etc before handing it over to the OS.
Yeah, I get that, RAM checks, a extension BIOS for every network card, the onboard RAID, the external RAID, the onboard SATA controller that isn't even used. I turn of as many of the unused ones as I can, but that still isn't going to get to quick.
And I don't need the server twiddling it's thumbs checking the RAID controllers battery when it was shut down cleanly for something routine like an OS update, or from the power management in a outage. I was already so sick of dumb RAID card behaviors that we started ripping them in favor of HBAs and letting the OS handle the storage stuff. Why wait at a text ROM screen watching as it scans your 18TB backup tier when you can boot the OS of mirrored SSDs and have your iSCSI targets up and running while you wait for a volume that won't get touched till end of day?
So actually, I think I DO have a problem with RAID cards that go blocking and/or have drivers so badly written they can't pause the volume and complete the consistency check after the OS loads. The manufactures are still building this stuff like it's the 1990s. How about disk recovery tools in case you need to load a disk set in another controller? Instead I get a RAID manager which only draws working buttons under IE8 and requires the java plugin?
If one of the Dell/HP hardware engineers reads this, you are seriously interfering with my ability to Mr Scott.
"Ditto for my switches, I have a stack of cisco 3850s that I swear take 15min to restart. "
Update your switches to a newer release - Cisco acknowledged the slow boot issue in IOS XE 3.x releases and put significant effort into reducing boot up times in the Denali (IOS XE 16.x) releases - should be down to less than 10 minutes for stacks of up to 6 switches (maybe more, I don't think my testing included larger stacks).
Add in switch priority config for all switches (primary/backup/members) and you cut off around an extra minute assuming all switches are powered up in a similar time frame.
And there's no front power switch due to a lack of real estate between cooling vents and ports for the PoE switches.
Thanks for the heads up on that!
I'm glad to see Cisco jump on the issue, and the broader firmware quality issues. We have been frozen on the same firmware version for a while, per our phone system integrator recommendations, as an earlier firmware had started silently dropping multicast traffic after running a few weeks. Kinda a problem when it silently dumps the emergency paging system on a campus in US, post parkland. After that we have been sticking to security fixes that had been vetted by the Integrator.
That said sawing 30% off my cold start times would be a big win. Cheers!
Per the current recommended version on the Cisco site: CSCvn30537
"After a LoadShed event, once the PS is restored, PoE is not restored back on the PD devices which went down."
Conditions: -Observed on a Stack using Stack Power.
Workaround: PoE is recovered only when the whole stack is power cycled (soft reboot doesn't recover)"
So the cure to my slow switch boot times may cause me to have to boot the whole stack twice, turning a 30% gain to a 30% loss.
Oh well, there might be build somewhere in the middle that isn't poison. Off to the patch notes. At least it's been 40 weeks since the last hard reboot.
Recently had my works PC changed to a cheap desktop but it has an SSD.
Boot times were fantastic but have slowed down a lot since all the usual Mcafee stuff has gone on. Still it's faster than Windows 7 - probably purely as it's a clean OS with fewer agents.
Always find that it's a case of the Windows folder bloating itself over time along with all the additional stuff that's needed to cover gaps in Windows in order to comply with standards - btw I'm not saying Linux wouldn't need additional software too, but we are talking about Windows.
They generally control plant equipment running into 6 figures, which is still working fine. The newer software only works with newer versions of the plant equipment, so a 6 figure reinvestment, because the OS is out of support.
Alternatively, just isolate the PCs from the network.
"What we are seeing right now is 99.9 per cent of all the apps that we are seeing on Windows 7 are compatible for Windows 10 and we can prove that with the data."
Ahh...statistics.
When you have 10,000 apps and 1000 don't work with Win10, discovering there are another 90,000 apps that work really helps the users of those non-working apps.
Someone's been sucking fitfully from the excuse vine.
I bought my son a 'gaming' laptop for his birthday and was incredibly disappointed to find that even with the Core i5 and 8GB RAM it takes him as long to start the laptop, login to windows and start his game as it used to take me when I was his age using an Amstrad 464.
Windows 10 has been getting steadily worse, and it wasn't exactly a great start point in first place.
> As many Reg readers know, Windows 10 devices can sometimes require users take a trip to the bathroom, make a fresh cup of coffee and have a stretch, before a usable cursor finally crawls onto the screen
Really? For me, Windows 10 boots are fairly quick and clean - the machine under the TV in the living room (steam/media streaming - I3, 8gb) gets rebooted pretty much daily, and is generally up and running in under 20 seconds, though it can take a bit longer for SMB mounts to become available.
The only time I debate nipping to put the kettle on is whenever Microsoft decides that there's some new patches which absolutely have to be installed.
Admittedly, it's a deliberately clean build and I've done my best to disable the various agents which Microsoft deem necessary - Skype, OneDrive, etc. But I suspect the issue is more about third party processes, plus the proliferation of underspecc'd machines - e.g. a Celeron with 4GB of ram; this would probably fly under Android, but Windows isn't quite as flexible.
(I have a Chuwei Hi12 12" dual-boot tablet - quad core atom, 4gb ram. It's noticeably more usable under Android as compared to Windows, though this may partly be due to the fact that no matter how much Microsoft pretends otherwise, Windows 10 just isn't fit for use in a tablet context...)
OTOH, back when I was in corporate land and using a laptop liberally festooned with HDD encryption and several dozen company-mandatory "agents" (AV, VPN, update monitors, etc), rebooting was actively painful regardless of which version of Windows was involved. Ironically, this often led to people forgoing reboots for as long as possible, resulting in larger exposure windows for zero-day exploits and the like!
I work in education and boot times on even the oldest Core2's is still around a minute or so at worst. They are all SSD drives but this biggest change is pruning the crap out of a W10 image before deploying it, as there is a mountain of unwanted shoite in there, even in the 'Education' version.
Time for the life support question again ... which may be asked of any Microsfot employee, but is best aimed at an executive or a particularly enthusiastic Windows marketurd:
"You are in hospital following a grievous accident or illness. There are two computers here. You'll get one of them. Only one of these computers, managing this amazing festoon of medical equipment around and inside you, will keep you alive for the next three months. The hardware and the medical software is excellent, but if the computer fails for any reason, and cannot be restarted to full functionality in a maximum of three or four minutes, you will slowly suffocate and die. So: would you prefer the one with Windows, or the one with the Linux operating system? Take your time answering ... the guy in the next bed gets the other one."
My wild-ass guess is that not a single Microsoft employee, nor anyone else with decent experience of both OSs would choose the Windows machine. Not where their life is at stake.
Well, would you?
I recently got an HP with an Intel I9 running at 4GHz, packed with 32Gb of 2.6GHz DDR4 RAM, & a 250Gb M.2 SSD.
I've disabled all the cruft I can find, uninstalled the crap AV they infested it with, & generally decrapified it as much as possible.
I told it to download the latest updates, let it apply them, & it reports itself as current.
It also takes minutes to finish jumping through hoops until it lets me have control.
I turn it on & go start the kettle, go to the loo, pour the water into the mug for a cuppa, then wait a few more seconds before it lets me do anything.
What the fuck am I doing wrong? Other than using Win10 on HP that is.
HP load up whole piles of crap onto their PCs. Maybe a clean install will help (though make sure you download any appropriate drivers first in case networking is borked when doing the clean install). Here's a few links that might help:
https://www.howtogeek.com/224342/how-to-clean-install-windows-10/
https://www.windowscentral.com/how-do-clean-installation-windows-10
https://www.groovypost.com/howto/clean-install-Windows-10/
Searching for "clean install windows 10" will give loads more.
Good luck!
if I could just be able to delay the forced updates on my work Win10 machine. When it decides it's going to update I get a 5 minute warning then it's going down. Last time this happened I was 45 minutes into a couple of 1 hour circuit tests that were going to ride me all the way to shift change. I managed to stop the tests, save what test data I had and get the test sets out of the circuits with about 15 seconds to spare. Wound up having to leave late because I didn't have time after the updates before end of shift to properly complete the ticket and hand it off for retest since the problem required a full hour of testing to prove the fix. The ability to just tell my computer "wait to update for X(up to 12) hours" by itself would be a huge help.
Depends of course on how locked down your admins have made your Win10 machine, but Microsoft have actually heard and answered this complaint. Go to Settings/Windows Update/Active hours and set the hours during which you don't want an automatic restart to occur. Add in a manual restart as your final action when you leave work and updates *shouldn't* interfere in future.
... and I am not kidding. My 5 years old ThinkPad only needs 15s to boot to login screen, or including the login (if I type the password fast enough) less than 20s to usable desktop (admittedly I switched it to UEFI boot to achieve that). On the other hand, Windows logins around here need more than 30s just to load user profile from the network, and that is on top of 3 minutes boot time and 15 minutes anti-virus scanner after boot which, although technically not a boot time, slows down the computer so much that it is unusable anyway.
Which is why I am so grateful to Steam for releasing more and more games on Linux - which means that my kids can use Linux for all of the things they use computer for.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2023
