A coarse grain overview of energy-aware resource management approaches since the last ten years through 2 EU founded projects. Presented during the 2014 "Energy Aware Network" Labex day at Inria. (see http://www.ucnlab.eu/fr/node/66)

1. energy efficient resource
management in virtualised
datacenters
Fabien Hermenier

2. USA 3 %
of the 2005 budget
(Environmental Protection Agency, 2007)
1.5%
in 2010 ?

3. 1Consume less
2003

4. the brown constant
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
180
175 W
170
160
150
140
130
120
110
0 1 2 3 4
VM #
Consumption (Watt)
Node consumption statistics (Cluster edel, 128GB HDD)
110 W

5. 2003elnozahy et al. :
“let’s turn off useless stuff”
Packing concentrate WWW
requests

6. 2006 Hansen et al. :
“look! look it’s moving”
live migration
N1 VM1 N2

7. consume less,
the theory
1) model power consumption
2) pack VMs
3) turn-off unused nodes
4) minimize(W)
5) profit

8. consume less,
the practice
hw. heterogeneity
env. capabilities
performance vs. energy
workload volatility
data center size
, ,

9. BtrPlace proposal
core reconfiguration
algorithm
users scripts
+ =
specialized reconfiguration
algorithm

10. spread(VM[2..3]);
preserve({VM1},’ucpu’, 3);
offline(@N4);
The reconfiguration plan
0’00
to
0’02:
relocate(VM2,N2)
0’00
to
0’04:
relocate(VM6,N2)
0’02
to
0’05:
relocate(VM4,N1)
0’04
to
0’08:
shutdown(N4)
0’05
to
0’06:
allocate(VM1,‘cpu’,3)
BtrPlace

11. the core reconfiguration
algorithm is modeled wrt. the
impact of actions on resources

12. Premadevariables,
spread({VM1,VM2}):
allDifferent(dhost
1 , dhost
2 ) ^
dhost
1 = chost
2 ! dst
1 # ced
2 ^
dhost
2 = chost
1 ! dst
2 # ced
1
constraints

13. Constraint Programming
!
!
!
to the rescue

14. 2010 2013
Energy aware ICT optimization policies
(+ btrPlace)

15. multi-core CPUs,
DDR3 memory,
spinning HD,
PUE / CUE,
boot/shutdown time
workload particularities
VM template
migration duration
migration payback time
hw. particularities
a fine-grain power model

16. energy-related variables are linked to
core ones

17. energy-oriented constraints
MaxServerPower
DelayBtwMigrations
DelayBtwServerSwitch
PayBackTime
SpareCPUs
minEnergyCons
minGasEmissions
cap consumption
reduce ping-pong effects
migration as an investment
control the consolidation aggressiveness
optimisation criteria

18. Time (minutes) P4G P4G + spare P4G + spare + vcpu P4G + spare + vcpu + delay
500 1000 1500 2000 2500
1
2
3
4
5
6
7
Servers
dominates scalability
the core problem

19. coarse to fine
grain optimisation
-16%
-47%
- 27%
-7%

20. Consuming less
dealing with
THE BROWN CONSTANT

21. Consuming less
dealing with
UNAPPROPRIATE
HARDWARE

22. energy proportional servers,
hw. community did not
chill
free cooling,
fanless processors

23. ? workload agnostic
need to revamp software approaches
migration
a balancing problem

24. Consume be2tter
2012

25. DC4Cities
2013 2016
let existing and new data centres become energy adaptive

26. align workload to renewable
energies availability

27. forecasts
1) 24h power
budget
2) alt. working
modes
energy adaptive
applications
3) selected mode
DC4Cities gray box approach

28. Energy adaptive
Batch scheduling
How many parallel jobs,
which ones to defer
Trade replicas
against latency
Allocate resources
against SLAs
Energy adaptive
WWW
Energy adaptive
IaaS
Energy adaptive
… …

29. ConclusionS

30. consume less, consume better, non-renewable
power sources, DVFS, live-migration,
deferrable workload, so many facets, non-deferrable
workload, elasticity, VM packing,
VOVO, VM balancing, white-box approach,
black-box approach, solutions need to follow
new capabilities and usage, priority over SLA,
priority over savings, renewable power sources,
fine grain power model, coarse grain power
model, steady workload, bursty workload

31. 1.1 - 1.5 %
in 2010
USA
(Environmental Protection Agency, 2013)

32. .org