On 11/12/2013 11:52 AM, Helmer Teles wrote:

On Tuesday 12 November 2013 11:31:38 Orion Poplawski wrote:

...

Is there any evidence that nicing something actually save energy? I wouldn't have thought so.

Hi, Orion.

With higher load comes higher CPU usage and with that comes more power consumption from the CPU, hardware, and Power Supply.

regards

But nicing something something just changes the scheduling, not the amount of work to be done.

-- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder/CoRA Office FAX: 303-415-9702 3380 Mitchell Lane orion@nwra.com Boulder, CO 80301 http://www.nwra.com

On 11/12/2013 12:06 PM, Helmer Teles wrote:

On Tuesday 12 November 2013 11:54:40 Orion Poplawski wrote:

...

But nicing something something just changes the scheduling, not the amount of work to be done.

But in my POV if you have a workload with 30 or more users running several applications, if the suspended processes can have a lower scheduling the CPU will have less wait states, but maybe you are right. :D

Powertop can help in this question ;)

Yes, I'd like to see empirical measurements - preferably with a sensitive watt meter :)

-- Orion Poplawski Technical Manager 303-415-9701 x222 NWRA, Boulder/CoRA Office FAX: 303-415-9702 3380 Mitchell Lane orion@nwra.com Boulder, CO 80301 http://www.nwra.com

Am 12.11.2013 20:15, schrieb Orion Poplawski:

On 11/12/2013 12:06 PM, Helmer Teles wrote:

...

On Tuesday 12 November 2013 11:54:40 Orion Poplawski wrote:

...

But nicing something something just changes the scheduling, not the amount of work to be done.

But in my POV if you have a workload with 30 or more users running several applications, if the suspended processes can have a lower scheduling the CPU will have less wait states, but maybe you are right. :D

Powertop can help in this question ;)

Yes, I'd like to see empirical measurements - preferably with a sensitive watt meter :)

I'm only guessing here, but while it is probably true, that a process with a lower priority will not decrease the power consumption of the server it is running on, you should see an increase in responsiveness for currently active tasks, like X2Go sessions in an active/connected state. Which would allow you to stack more users onto a single server, rather than having to buy a second one. And that's quite an energy consumption there, both in electricity during its lifetime and in electricity/resources needed to build the server. Of course, this is assuming that a certain percentage of your users is always idle/suspended/disconnected or not logged in at all.

-Stefan

Hello!

The renicer just adjust the execution priority of a suspended session. Where you'd see a noticeable impact would be in a situation where you have a lot of suspended users. Which may be the normal case in a academic situation where you have significantly more user accounts than termial seats. Lets say you have 10-20 active terminals and 50-100 suspended users. By dropping the priority of the suspended sessions to rock bottom, the server will do everything else first... then tend to the suspended session on its "spare time".

As some have mentioned already... By it self, on any given hardware, it does not save power... But it does increase the amount of actively suspended users you can run on that hardware (while still maintaining an acceptable performance), thereby reducing the hardware requirement.

But together with CPU throttling based on how many active sessions you have and what those users may be doing... you do save power...

In a traditional server environment you'd probably benefit from the renicer... however getting into CPU throttling and stuff like that to shave a few watts of your consumption is not really worth it.

but when your trying to build a terminal server with a 50-100w power envelope... and you want this particular server to serve 10-20 active terminals. And you're wanting to be able to run the whole system on the smallest possible solar power setup... That's when you need to get creative.

Every saved watt * daily run time = significant impact on power system requirement and cost.

Best Regards, -GZNG

On 13. nov. 2013 08:29, Mike Gabriel wrote:

Hi all,

nice discussion going on...

On Di 12 Nov 2013 19:31:38 CET, Orion Poplawski wrote:

...

On 11/05/2013 09:12 AM, Mike Gabriel wrote:

...

Hi all,

the 4.1.0.0-preview (x2goserver.git master) just received a code patch from Guangzhou Nianguan Electronics Technology Co.Ltd, an open source company in Asia that plans to use X2Go for school setups in areas where electricity is a rare resource. They plan to build cost-effective and energy-effective embedded thin clients running X2Go.

The patch we received adds a feature to X2Go Server called SupeReNicer [1] and it is a tool that hooks into x2gocleansessions. It renices suspended sessions to nice level 19 and renices them again to nice level 0 once they change there status from suspended to running.

Is there any evidence that nicing something actually save energy? I wouldn't have thought so.

The intention of the renicing code is rather about squeezing more users on one server than reducing power consumption.

If anyone indeed does some power measuring, please keep us in the loop about your results.

Mike

---

X2Go-Dev mailing list X2Go-Dev@lists.berlios.de https://lists.berlios.de/mailman/listinfo/x2go-dev