The graphs in the Linode Manager are excellent for looking for CPU, IO or networking peaks on my own Linode. Some of the peaks are out of my control, but some are actually triggered by daily cron jobs doing web statistics, backups and so on.

Now, I'd like to place these IO and CPU intensive tasks on ideal timeslots during the night, so that the effect for other Linodes on the machine is reduced. Also, my own backup jobs might complete a bit quicker (always good).

Currently, however, I have no way of knowing what the host machine CPU and IO graphs look like (to find reasonable slots for my jobs). I understand that this might be seen as a security/privacy/whatnot concern, but I think it would be useful also in other situations. To find the underlying reasons for temporary slow performance, for instance.

Anybody else that would see this a useful? Or are there already public graphs of this available somewhere?

I second this. Even if displaying the host graphs is not an option (for whatever reason) some sort of indication about host load (other than the "current load" on the dashboard) would be nice. Maybe just a simple string like:

Makes sense?

Perhaps the XML API could have a "host nice" feature, rather than everyone scheduling their disk intensive jobs for the same lull every evening.

I think in the XEN enviroment that this wouldnt make a lot of sense, atleast not from the CPU usage. When you think about the fact that you have access to 4 out of 8 cores.. and i assume the distribution is somewhat managed.. ie
Person a has access to cores 1,2,3,4
Person b has access to cores 5,6,7,8
Person c has access to cores 1,2,5,6
Person c has access to cores 3,4,7,8

I assume you get multiple permutations of this agsinst the x number of people on the host. Then the fact that at 10pm the server shows there is a spike in usage, might be totally irrelevant to you because you dont have access to any of the cores that are spiking.

Host load IO and Networking graphs would be useable though as we all share the same disk io and network connections..

_________________
ServerAdmin - www.our-lan.com
"Diplomacy is the art of saying nice doggy whilst looking for a really big stick"
"In my experiece, any attempt to make any system idiot proof will only challenge God to make a better idiot"

But if it is "managed" as you said, wouldn't make it easier? An example might be:

What I mean is if there's any kind of pattern regarding cpu distrubition among linodes, then it's "theoratically" possible to obtain some info about it.

Well..
Assuming my math is some what correct, the number of possible permutations is 8P4, which is
P(8,4) = 8! / (8 - 4)! = 1680

which would indicate there is 1680 different combinations.. since linodes have a max of 40 nodes per host.. In theory no one may have the same combination as you.

I cant speak for the linode admins, but i would be interested to know how they do alocate the cores.. If it is just a few groups and a roundrobin approach to which selection of cores you get.. or if they balance it out some other wya..

_________________
ServerAdmin - www.our-lan.com

"Diplomacy is the art of saying nice doggy whilst looking for a really big stick"

"In my experiece, any attempt to make any system idiot proof will only challenge God to make a better idiot"

It's not a permutation, it's a combination.

It doesn't matter if I'm on cores 1, 4, 2, and 7 or 7, 4, 2, and 1

(Your math is still correct, I'm just being pedantic)

Oh.. my bad.
You are entirely correct. It should be combinations not permutations..
In which case its 8C4, or (8!)/(4!(8-4)!) = 70.

Still more combinations then a linode can have

_________________
ServerAdmin - www.our-lan.com

"Diplomacy is the art of saying nice doggy whilst looking for a really big stick"

"In my experiece, any attempt to make any system idiot proof will only challenge God to make a better idiot"

I guess most people will schedule their CPU/IO-heavy cron jobs at something like 2:00AM or 4:00AM, so I avoid those X o'clocks and schedule my cron jobs at odd minutes like 3:37AM or 5:19AM.

No useful cron job should hog CPU and IO for more than a few minutes a day anyway.

Also note that not everyone uses US timezone...

I am scheduling my offsite backup + recalculation of data in early morning (2AM to 4AM), but in Australian Eastern Standard Time (+10 + DST at the moment), which would be around 7AM to 9AM Pacific time for my Fremont Linode (and 9AM to 11AM during your summer)...