Information Technology Reference
In-Depth Information
full capacity, the energy when not all the available cores of the machine are used,
and the energy that each machine consumes in idle state (
MIN-MAX
mode).
Iturriaga
et al.
[11] showed that a parallel multiobjective local search based on
Pareto dominance outperforms deterministic heuristics based on the traditional
Min-Min strategy. We also apply the MIN-MAX mode in this present article.
Dorronsoro
et al.
[6] presented a two-level strategy for scheduling large par-
allel applications in multicore distributed systems, minimizing the total compu-
tation time and the energy consumption. The approach combines a higher-level
(between distributed datacenters) and a lower-level (within each datacenter)
scheduler. Accurate schedules were computed by using heuristics accounting for
both problem objectives in the higher level, and ad-hoc schedulers to take ad-
vantage of multicore infrastructures in the lower level. We adapt three low-level
schedulers from that previous article for the problem we consider in this work.
We initially explored the application of multiobjective control planning for
datacenters in [25]. We introduced the problem model and initial results about
controlling datacenter power consumption while maintaining temperature and
QoS levels. The present work extends the previous study, focusing on further an-
alyzing the multiobjective approach and studying different scheduling strategies
for reducing power consumption and increasing QoS.
3 Control and Scheduling Approach for Energy-Aware
Datacenters
This section describes the approach for control and energy-aware scheduling in
datacenters, according to reference power consumption profiles.
3.1 Datacenter Model
Fig. 1 shows the datacenter model applied in this work. We follow the power man-
agement approach applied to the Parasol datacenter [8], considering two systems:
heating-ventilation-air conditioning (HVAC) and computing infrastructure (IT).
The
control signals
are variables that alter the behavior of the datacenter:
1. HVAC control (
c
k
): we consider a datacenter equipped with free and AC
cooling infrastructure.
c
k
comprises a set of signals: AC compressor state
(binary), free cooling fan speed (%), and free cooling damper state (binary).
2. Schedule (
s
k
): shape the IT power consumption, considering the number of
servers running, load constraints, and specific user requirements.
The
controllable variables
are handled via manipulation of the control signals:
1. Quality of service (
QoS
k
): depends on user and system related metrics, which
are computed using a specific scheduling strategy applied in the datacenter.
2. Internal temperature (
T
k
): thermostat reading of the datacenter (
o
C
).
3. Cooling power (
C
k
): the sum of AC power and free cooling fan power (kW).
4. IT power (
I
k
): the power of the servers, switches and all IT equipment in
the datacenter (kW).
5. Total power (
P
k
): the total power used by the datacenter (
P
k
=
C
k
+
I
k
).
