WSUS Client download timing

The subject of when the client downloads is often discussed here but one thing I
wonder about is the effect of daily shutdowns.

Normally the Detection is every 17 to 22 hours so if the PCs are always on the
detections will eventually spread out throughout the whole 24 hours.

If a PC is switched off when the next detection is due it will run a detection
on boot up. The downloads will happen a few minutes after the boot up (as soon
as detection finishes).

So if all the PCs are switched off every night and are all switched on in the
morning at the same time (more or less) will this result in each PC starting the
download at the same time. That is they are all off at the time the detection is
scheduled which would be between 00:00 and 07:00. Then at 08:00 all the PCs are
switched on and see the missed detection. So just after 08:00 the whole lot
complete a detection and start a download.

If the above is a correct interpretation then switching off every night will
force the WSUS downloads to concentrate at a time just around company opening
hours.

Just wondering if I have worked this out correctly or is there some other factor
I missed which will prevent this happening. If it is so than a default
detection cycle of 24 to 30 hours would be better since a PC will normally be
switched on at about the same time each day so will not yet have passed the
scheduled detection time. The detection will then not happen until the scheduled
time and thus with the random offset will continue to spread out throughout the
day.
--
Dave Mills
There are 10 types of people, those that understand binary and those that don't.

"Dave Mills" <> wrote in message
news:...
> The subject of when the client downloads is often discussed here but one
> thing I
> wonder about is the effect of daily shutdowns.
>
> Normally the Detection is every 17 to 22 hours so if the PCs are always on
> the
> detections will eventually spread out throughout the whole 24 hours.
>
> If a PC is switched off when the next detection is due it will run a
> detection
> on boot up. The downloads will happen a few minutes after the boot up (as
> soon
> as detection finishes).
>
> So if all the PCs are switched off every night and are all switched on in
> the
> morning at the same time (more or less) will this result in each PC
> starting the
> download at the same time.

Theoretically..., but luckily your scenario is not quite this likely.

> That is they are all off at the time the detection is
> scheduled which would be between 00:00 and 07:00.

With a normal detection interval of 22 hours, if all machines are powered
down between midnight and 7am, then, statistically speaking, only 1/3 of the
clients will actually miss their scheduled installation. They will execute a
detection at power on, and the next detection will then be scheduled for
17-22 hours later. This cycle will repeat, and all detections will function
normally, until that 17-22 hour recycle time puts the next detection
somewhere between midnight and 7am.
It's likely this scenario will repeat every couple of days with a different
group of 1/3 of the total number of computers.

> If the above is a correct interpretation then switching off every night
> will
> force the WSUS downloads to concentrate at a time just around company
> opening
> hours.

Well, at least 1/3 of the systems -- yes -- and this is a good place to
point out that the design criteria assumes that computers are powered on
24x7, and has capabilities built in to handle the exceptions, but not to the
extent of an entire organization being powered down on a routine basis.

> Just wondering if I have worked this out correctly or is there some other
> factor
> I missed which will prevent this happening. If it is so than a default
> detection cycle of 24 to 30 hours would be better since a PC will normally
> be
> switched on at about the same time each day so will not yet have passed
> the
> scheduled detection time. The detection will then not happen until the
> scheduled
> time and thus with the random offset will continue to spread out
> throughout the
> day.

Except that 22 hours is the maximum configurable detection interval. But
even if you could set it to >24 hours, the problem will still occur because
of the -1 to -20% offset in the detection interval, which is designed to
ensure that client detections are spread out around the clock, and thus
balance the load on the WSUS Server.

At some point every PC will cycle through to an after-midnight detection
about every seven to ten days or so.


--
Lawrence Garvin, M.S., MCITP:EA, MCDBA
Principal/CTO, Onsite Technology Solutions, Houston, Texas
Microsoft MVP - Software Distribution (2005-2009)

My Blog: http://onsitechsolutions.spaces.live.com
Microsoft WSUS Website: http://www.microsoft.com/wsus
My MVP Profile: http://mvp.support.microsoft.com/pro...awrence.Garvin

On Wed, 23 Sep 2009 07:54:18 -0500, "Lawrence Garvin [MVP]"
<> wrote:

>"Dave Mills" <> wrote in message
>news:.. .
>> The subject of when the client downloads is often discussed here but one
>> thing I
>> wonder about is the effect of daily shutdowns.
>>
>> Normally the Detection is every 17 to 22 hours so if the PCs are always on
>> the
>> detections will eventually spread out throughout the whole 24 hours.
>>
>> If a PC is switched off when the next detection is due it will run a
>> detection
>> on boot up. The downloads will happen a few minutes after the boot up (as
>> soon
>> as detection finishes).
>>
>> So if all the PCs are switched off every night and are all switched on in
>> the
>> morning at the same time (more or less) will this result in each PC
>> starting the
>> download at the same time.
>
>Theoretically..., but luckily your scenario is not quite this likely.
>
>> That is they are all off at the time the detection is
>> scheduled which would be between 00:00 and 07:00.
>
>With a normal detection interval of 22 hours, if all machines are powered
>down between midnight and 7am, then, statistically speaking, only 1/3 of the
>clients will actually miss their scheduled installation. They will execute a
>detection at power on, and the next detection will then be scheduled for
>17-22 hours later. This cycle will repeat, and all detections will function
>normally, until that 17-22 hour recycle time puts the next detection
>somewhere between midnight and 7am.
>It's likely this scenario will repeat every couple of days with a different
>group of 1/3 of the total number of computers.

Are you therefore saying that the next detection will be 17-22 hours after the
last scheduled detection and not 17-22 hours after the current detection. I have
not looked for that, I had not even thought of it.

If so you are correct. But if the next detection is 17-22 after the current
(delayed) detection then all the PCs will migrate to the same cycle.

I need to have a look for this. None of my lap PCs ever get shut down so I will
see nothing there.

>
>> If the above is a correct interpretation then switching off every night
>> will
>> force the WSUS downloads to concentrate at a time just around company
>> opening
>> hours.
>
>Well, at least 1/3 of the systems -- yes -- and this is a good place to
>point out that the design criteria assumes that computers are powered on
>24x7, and has capabilities built in to handle the exceptions, but not to the
>extent of an entire organization being powered down on a routine basis.

Which is the norm for a school where the geography department goes around
sticking "save energy - switch it off" stickers on everything.

>
>> Just wondering if I have worked this out correctly or is there some other
>> factor
>> I missed which will prevent this happening. If it is so than a default
>> detection cycle of 24 to 30 hours would be better since a PC will normally
>> be
>> switched on at about the same time each day so will not yet have passed
>> the
>> scheduled detection time. The detection will then not happen until the
>> scheduled
>> time and thus with the random offset will continue to spread out
>> throughout the
>> day.
>
>Except that 22 hours is the maximum configurable detection interval.

You are simply stating the current design limitation. There is no reason that
this cannot be changed.

>But
>even if you could set it to >24 hours, the problem will still occur because
>of the -1 to -20%

Another current design limitation but there is no reason the offset could not be
+1% to +20% (although +20% would be too big)


>offset in the detection interval, which is designed to
>ensure that client detections are spread out around the clock, and thus
>balance the load on the WSUS Server.
>
>At some point every PC will cycle through to an after-midnight detection
>about every seven to ten days or so.
--
Dave Mills
There are 10 types of people, those that understand binary and those that don't.

Dave Mills wrote:

> The subject of when the client downloads is often discussed here but one thing I
> wonder about is the effect of daily shutdowns.

OK, I'm going to suppose that the machines are normally on from 9am to 5pm.
Taking away 17 hours from 9am gives us 3pm, so if all the machines were to
detect between 3pm and 5pm, they'd next detect between 9am and 3pm (the latter
being 5pm plus 22 hours).

You'd still be doing a detection on every machine over the two-hour interval
from 3pm to 5pm, about ten times the design load, but this might still be OK
depending on how many machines we're talking about and how powerful the WSUS
server is.

You could schedule these detections using the AT command to run wuauclt
/detectnow, or using scripting, or in various other ways - the usual suspects.

Harry.

>
> Normally the Detection is every 17 to 22 hours so if the PCs are always on the
> detections will eventually spread out throughout the whole 24 hours.
>
> If a PC is switched off when the next detection is due it will run a detection
> on boot up. The downloads will happen a few minutes after the boot up (as soon
> as detection finishes).
>
> So if all the PCs are switched off every night and are all switched on in the
> morning at the same time (more or less) will this result in each PC starting the
> download at the same time. That is they are all off at the time the detection is
> scheduled which would be between 00:00 and 07:00. Then at 08:00 all the PCs are
> switched on and see the missed detection. So just after 08:00 the whole lot
> complete a detection and start a download.
>
> If the above is a correct interpretation then switching off every night will
> force the WSUS downloads to concentrate at a time just around company opening
> hours.
>
> Just wondering if I have worked this out correctly or is there some other factor
> I missed which will prevent this happening. If it is so than a default
> detection cycle of 24 to 30 hours would be better since a PC will normally be
> switched on at about the same time each day so will not yet have passed the
> scheduled detection time. The detection will then not happen until the scheduled
> time and thus with the random offset will continue to spread out throughout the
> day.

"Dave Mills" <> wrote in message
news:...

>>With a normal detection interval of 22 hours, if all machines are powered
>>down between midnight and 7am, then, statistically speaking, only 1/3 of
>>the
>>clients will actually miss their scheduled installation. They will execute
>>a
>>detection at power on, and the next detection will then be scheduled for
>>17-22 hours later. This cycle will repeat, and all detections will
>>function
>>normally, until that 17-22 hour recycle time puts the next detection
>>somewhere between midnight and 7am.
>>It's likely this scenario will repeat every couple of days with a
>>different
>>group of 1/3 of the total number of computers.
>
> Are you therefore saying that the next detection will be 17-22 hours after
> the
> last scheduled detection and not 17-22 hours after the current detection.

That also depends. If a scheduled detection was successful *prior* to the
machine being powered off, and the machine is powered back on *before* the
next scheduled detection, then the machine will not perform a detection at
power on.

Furthermore, if a scheduled detection was supposed to occur during the time
the machine was powered off, a detection will occur at power on and the
*next* detection will be scheduled for 17-22 hours after the power-on
detection, not the missed detection.

> If so you are correct. But if the next detection is 17-22 after the
> current
> (delayed) detection then all the PCs will migrate to the same cycle.

Mathematically impossble. :-)

>>> Just wondering if I have worked this out correctly or is there some
>>> other
>>> factor
>>> I missed which will prevent this happening. If it is so than a default
>>> detection cycle of 24 to 30 hours would be better since a PC will
>>> normally
>>> be
>>> switched on at about the same time each day so will not yet have passed
>>> the
>>> scheduled detection time. The detection will then not happen until the
>>> scheduled
>>> time and thus with the random offset will continue to spread out
>>> throughout the
>>> day.
>>
>>Except that 22 hours is the maximum configurable detection interval.
>
> You are simply stating the current design limitation. There is no reason
> that
> this cannot be changed.

Good luck with that! <vbg>

>
>>But
>>even if you could set it to >24 hours, the problem will still occur
>>because
>>of the -1 to -20%
>
> Another current design limitation but there is no reason the offset could
> not be
> +1% to +20% (although +20% would be too big)

And not likely to be implemented any time soon, either. ;-)

But, yes, *theoretically* that could be done.

Practically speaking, one could purchase a license for the EminentWare WSUS
Extension Pack and discretely schedule each client's detections at exactly
the time desired every day, and completely avoid the possibility that
detections would fall during power-off hours, or that any detections would
ever occur at power on.


--
Lawrence Garvin, M.S., MCITP:EA, MCDBA
Principal/CTO, Onsite Technology Solutions, Houston, Texas
Microsoft MVP - Software Distribution (2005-2009)

My Blog: http://onsitechsolutions.spaces.live.com
Microsoft WSUS Website: http://www.microsoft.com/wsus
My MVP Profile: http://mvp.support.microsoft.com/pro...awrence.Garvin

In article <>, News1@nospam--
djmills-dot-co.uk says...
<snip>
> So if all the PCs are switched off every night and are all switched on in the
> morning at the same time (more or less) will this result in each PC starting the
> download at the same time. That is they are all off at the time the detection is
> scheduled which would be between 00:00 and 07:00. Then at 08:00 all the PCs are
> switched on and see the missed detection. So just after 08:00 the whole lot
> complete a detection and start a download.
<snip>
This scenario is similar to what happens in my WSUS environment.
Lawrence pointed out that 1/3 of the PC's will get updated in the
morning. After I approve the updates it usually takes 2 or 3 days to
fully install all the patches throughout the building.

The reboot is forced so the reboot message occurs about 15 to 20 minutes
at the start of the day. Causing some feedback by some of the users, but
we have been operating like this for a few years. I think the users are
getting used to the reboots.
--
antwesor

"State of MN" <> wrote in message
news: ...

> After I approve the updates it usually takes 2 or 3 days to
> fully install all the patches throughout the building.
>
> The reboot is forced so the reboot message occurs about 15 to 20 minutes
> at the start of the day. Causing some feedback by some of the users, but
> we have been operating like this for a few years. I think the users are
> getting used to the reboots.

What would be their reaction if they found out they could fix the issue by
simply leaving machines powered on and moving the scheduled installation
time (with unattended restart) to overnight hours -- all for the low low
price of a few watt-hours of electricity once a month?

A whole thousand PCs can be updated once-a-month, overnight, for less than a
buck a month. I doubt the lost productivity for those morning
installs/restarts can be measured in the "less than a buck a month" range.


--
Lawrence Garvin, M.S., MCITP:EA, MCDBA
Principal/CTO, Onsite Technology Solutions, Houston, Texas
Microsoft MVP - Software Distribution (2005-2009)

My Blog: http://onsitechsolutions.spaces.live.com
Microsoft WSUS Website: http://www.microsoft.com/wsus
My MVP Profile: http://mvp.support.microsoft.com/pro...awrence.Garvin

State of MN wrote:

> The reboot is forced so the reboot message occurs about 15 to 20 minutes
> at the start of the day. Causing some feedback by some of the users, but
> we have been operating like this for a few years. I think the users are
> getting used to the reboots.

The updates might all detect in the morning, but why would they install straight
away? Are you setting deadlines?

It would seem more sensible to set things up so that the "install on shutdown"
option is available, if the machines are being shut down each night anyway.

Harry.

"Harry Johnston [MVP]" <> wrote in message
news:...
> State of MN wrote:
>
>> The reboot is forced so the reboot message occurs about 15 to 20 minutes
>> at the start of the day. Causing some feedback by some of the users, but
>> we have been operating like this for a few years. I think the users are
>> getting used to the reboots.
>
> The updates might all detect in the morning, but why would they install
> straight away?

A default scheduled installation event at 3am, which is not allowed to
execute due to the machine being powered off, combined with the default
setting to install updates at one minute past service startup for missed
scheduled installations would cause exactly this behavior.

> It would seem more sensible to set things up so that the "install on
> shutdown" option is available, if the machines are being shut down each
> night anyway.

Absolutely! And combine that with moving the scheduled installation event to
sometime shortly after normal quitting time, so that those who leave their
machines logged on/powered up, will have the updates installed before
shutdown, and be prompted to restart/shutdown (until they actually do, or
log off without shutting down).

Note: The Install Updates and Shutdown option is available, by default. The
policy would have to be explicitly set to "Enabled" (it's a reverse
semantics policy setting) in order to prevent access to that feature.

--
Lawrence Garvin, M.S., MCITP:EA, MCDBA
Principal/CTO, Onsite Technology Solutions, Houston, Texas
Microsoft MVP - Software Distribution (2005-2009)

My Blog: http://onsitechsolutions.spaces.live.com
Microsoft WSUS Website: http://www.microsoft.com/wsus
My MVP Profile: http://mvp.support.microsoft.com/pro...awrence.Garvin

In article <B16E236B-2182-47EB-BC62->,
ias says...
> "State of MN" <> wrote in message
> news: ...
<snip>
> What would be their reaction if they found out they could fix the issue by
> simply leaving machines powered on and moving the scheduled installation
> time (with unattended restart) to overnight hours -- all for the low low
> price of a few watt-hours of electricity once a month?
>
> A whole thousand PCs can be updated once-a-month, overnight, for less than a
> buck a month. I doubt the lost productivity for those morning
> installs/restarts can be measured in the "less than a buck a month" range.
>
I agree totally with your cost analysis of the situation, but the
policies are dictated from the commissioners office. I have tried to
explain to the management but to no avail. -sigh-
--
antwesor