Browse Prior Art Database

Method and System for Computing Power-Consumption of Each Server in a Server-Collection

IP.com Disclosure Number: IPCOM000193334D
Original Publication Date: 2010-Feb-19
Included in the Prior Art Database: 2010-Feb-19
Document File: 6 page(s) / 158K

Publishing Venue

IBM

Abstract

A method and system for computing power-consumed and cooling-demand by each server in a server-collection by utilizing a monitoring system is disclosed.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 47% of the total text.

Page 1 of 6

Method and System for Computing Power -Consumption of Each Server in a Server-Collection

Disclosed is a method and system for computing power-consumed and cooling-demand by each server in a server-collection by utilizing a monitoring system. The method involves computing

powe

system uses multiple inputs for the computation. These inputs may include list of the servers in the collection (i.e. server-collection model), current data about CPU utilization metrics of each server in the collection, current total-

powe

total-cooling-metrics of the collection. The monitoring unit also uses historical data, such as historical data about CPU utilization metrics of each server in the collection, historical total-

r-metrics of the collection, and historical total-cooling-metrics of the collection.

A topology of the monitoring system with monitoring agents & collector-component is illustrated in figure.

Figure

The method involves configuring a server cooling collector and a server power collector with the server-collection model which comprises information about number of servers in collection and identity of each server {S1, S2,… S

}.

Thereafter, the server cooling collector and the server power collector collect various metrics from a cooling-device agent and a power-device agent. One of these metrics is CPU Utilization metrics of each server i.e. x(i, k), which is CPU utilization of the server Si at time tk. Other metrics collected are total-

, which is total-

the collection at time tk and total-cooling-demand by the collection i.e. zk, which is total-cooling-demand by the collection at time tk. These metrics are stored in respective historical-data-cache.

Based on these metrics, dynamic power-model and dynamic cooling-model are computed. This

r-consumed and cooling-demand by each server in a collection of servers. The monitoring

r-metrics of the collection and current

powe

N

powe

powe

r-consumed by the collection i.e. y

r-consumed by

k

1

[This page contains 1 picture or other non-text object]

Page 2 of 6

is accomplished by sampling entries (M) from the historical-data-cache maintained by the Server Cooling Collector and the Server Power Collector such that, the CPU utilization of all the servers in the collection operates at different levels in the range 0 to 100, in any given sample. Further, the CPU utilization of all the servers in the collection is spread widely across the range 0 to 100 across multiple samples. Computation of the power model for each server is then performed using equations 1-4, to determine the regression coefficients a0, a1, a2, a3, etc.

N

N

N

MNa

0 K

1 (1)

a

+1∑∑ =

= =

M

k

xi y

x

a

+2∑∑

M

k

x

a

+3∑∑

(

k

M

k

=

,

)

(

i

,

k

)

(

i

,

k

)

M

k

k

= =

1 1

i

= =

1 1

2

i

1 1

3

i

N

N

×

N

N

×

N

∑∑

Na

M

k

x

a

+1∑ ∑

x

x

a

2∑ ∑

×

x

o

(

i

,

k

)

M

k

×

(

i

,

)

+

x

k

(

i

,

k

M

k

(

i

,

k

...