Worth their watts? - an empirical study of datacenter servers

Vasan, Arunchandar; Sivasubramaniam, Anand; Shimpi, Vikrant; Sivabalan, T.; Subbiah, Rajesh

doi:10.1109/hpca.2010.5463056

Cited by 72 publications

(45 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Using SPECjbb, we construct four representative power demand curves (P demand (t)) that are shown in Figure 6(a). Our demand curves are designed to represent the diverse peak characteristics that we observe with four type of realworld workloads: (i) 'Flash' has tall and narrow (15 minutes) peaks that are caused by flash crowds seen by many videostreaming/e-commerce applications [30], (ii) 'TCS' has moderately tall peaks lasting a little longer (1 hour), corresponding to a production datacenter workload of TCS, an IT services company [44], (iii) 'Google' has shorter but wider (3 hours) peaks similar to the Google cluster trace [23], and, (iv) 'MSN' has tall and very wide (8 hours) peaks similar to the Microsoft messenger workload [9]. We capture these different power characteristics in our demand curves by varying the SPECjbb workload intensity every b=15-minutes at all servers over a H=1 day window.…”

Section: Our Experimental Platformmentioning

confidence: 99%

Benefits and limitations of tapping into stored energy for datacenters

Govindan

Sivasubramaniam

Urgaonkar

2011

Proceedings of the 38th Annual International Symposium on Computer Architecture

Self Cite

153

View full text Add to dashboard Cite

Abstract. Datacenter power consumption has a significant impact on both its recurring electricity bill (Op-ex) and one-time construction costs (Cap-ex). Existing work optimizing these costs has relied primarily on throttling devices or workload shaping, both with performance degrading implications. In this paper, we present a novel knob of energy buffer (eBuff) available in the form of UPS batteries in datacenters for this cost optimization. Intuitively, eBuff stores energy in UPS batteries during "valleys" -periods of lower demand, which can be drained during "peaks" -periods of higher demand. UPS batteries are normally used as a fail-over mechanism to transition to captive power sources upon utility failure. Furthermore, frequent discharges can cause UPS batteries to fail prematurely. We conduct detailed analysis of battery operation to figure out feasible operating regions given such battery lifetime and datacenter availability concerns. Using insights learned from this analysis, we develop peak reduction algorithms that combine the UPS battery knob with existing throttling based techniques for minimizing datacenter power costs. Using an experimental platform, we offer insights about Op-ex savings offered by eBuff for a wide range of workload peaks/valleys, UPS provisioning, and application SLA constraints. We find that eBuff can be used to realize 15-45% peak power reduction, corresponding to 6-18% savings in Op-ex across this spectrum. eBuff can also play a role in reducing Cap-ex costs by allowing tighter overbooking of power infrastructure components and we quantify the extent of such Cap-ex savings. To our knowledge, this is the first paper to exploit stored energy -typically lying untapped in the datacenter -to address the peak power draw problem.

show abstract

Section: Our Experimental Platformmentioning

confidence: 99%

Benefits and limitations of tapping into stored energy for datacenters

Govindan

Sivasubramaniam

Urgaonkar

2011

Proceedings of the 38th Annual International Symposium on Computer Architecture

Self Cite

153

View full text Add to dashboard Cite

show abstract

“…The actual value of P sw usually depends on the technical specification of the servers (number of processors, number of cores per processor, memory, I/O devices). Server power consumption values are in the range of 200-600 W based on a measurement study of twenty different production data center servers [22]. The switches are modeled after the Arista 7504 switch, which is 7 RU high and has 192 SFP+ ports.…”

Section: A Rack Power Densitymentioning

confidence: 99%

Towards 48-fold cabling complexity reduction in large flattened butterfly networks

Csernai¹,

Ciucu

Braun

et al. 2015

2015 IEEE Conference on Computer Communications (INFOCOM)

View full text Add to dashboard Cite

“…We start with those not considering frequency. In this category we find articles proposing models where server components follow a linear behavior like [11,14,19] or more complex ones, like in [2,5,13]. In [14], Liu et al propose a simple linear model and evaluate different hardware configurations and types of workloads by varying the number of available cores, the available memory, and considering also the contribution of other components such as disks.…”

Section: Related Workmentioning

confidence: 99%

“…In [14], Liu et al propose a simple linear model and evaluate different hardware configurations and types of workloads by varying the number of available cores, the available memory, and considering also the contribution of other components such as disks. Vasan et al [19] monitored multiple servers on a datacenter as well as the power consumption of several of the internal elements of a server. However, they considered that the behavior of this server could be approximated by a model based only on CPU utilization.…”

Section: Related Workmentioning

confidence: 99%

A measurement-based analysis of the energy consumption of data center servers

Aroca

Chatzipapas

Anta

et al. 2014

Proceedings of the 5th International Conference on Future Energy Systems

View full text Add to dashboard Cite

Energy consumption is a growing issue in data centers, impacting their economic viability and their public image. In this work we empirically characterize the power and energy consumed by different types of servers. In particular, in order to understand the behavior of their energy and power consumption, we perform measurements in different servers. In each of them, we exhaustively measure the power consumed by the CPU, the disk, and the network interface under different configurations, identifying the optimal operational levels. One interesting conclusion of our study is that the curve that defines the minimal CPU power as a function of the load is neither linear nor purely convex as has been previously assumed. Moreover, we find that the efficiency of the various server components can be maximized by tuning the CPU frequency and the number of active cores as a function of the system and network load, while the block size of I/O operations should be always maximized by applications. We also show how to estimate the energy consumed by an application as a function of some simple parameters, like the CPU load, and the disk and network activity. We validate the proposed approach by accurately estimating the energy of a map-reduce computation in a Hadoop platform.

show abstract

Worth their watts? - an empirical study of datacenter servers

Cited by 72 publications

References 16 publications

Benefits and limitations of tapping into stored energy for datacenters

Benefits and limitations of tapping into stored energy for datacenters

Towards 48-fold cabling complexity reduction in large flattened butterfly networks

A measurement-based analysis of the energy consumption of data center servers

Contact Info

Product

Resources

About