Anshul Gandhi

Optimal power allocation in server farms

Server farms today consume more than 1.5% of the total electricity in the U.S. at a cost of nearly

AutoScale: Dynamic, Robust Capacity Management for Multi-Tier Data Centers

4.5 billion. Given the rising cost of energy, many industries are now seeking solutions for how to best make use of their available power. An important question which arises in this context is how to distribute available power among servers in a server farm so as to get maximum performance. By giving more power to a server, one can get higher server frequency (speed). Hence it is commonly believed that, for a given power budget, performance can be maximized by operating servers at their highest power levels. However, it is also conceivable that one might prefer to run servers at their lowest power levels, which allows more servers to be turned on for a given power budget. To fully understand the effect of power allocation on performance in a server farm with a fixed power budget, we introduce a queueing theoretic model, which allows us to predict the optimal power allocation in a variety of scenarios. Results are verified via extensive experiments on an IBM BladeCenter. We find that the optimal power allocation varies for different scenarios. In particular, it is not always optimal to run servers at their maximum power levels. There are scenarios where it might be optimal to run servers at their lowest power levels or at some intermediate power levels. Our analysis shows that the optimal power allocation is non-obvious and depends on many factors such as the power-to-frequency relationship in the processors, the arrival rate of jobs, the maximum server frequency, the lowest attainable server frequency and the server farm configuration. Furthermore, our theoretical model allows us to explore more general settings than we can implement, including arbitrarily large server farms and different power-to-frequency curves. Importantly, we show that the optimal power allocation can significantly improve server farm performance, by a factor of typically 1.4 and as much as a factor of 5 in some cases.

Minimizing data center SLA violations and power consumption via hybrid resource provisioning

Energy costs for data centers continue to rise, already exceeding

Are sleep states effective in data centers

15 billion yearly. Sadly much of this power is wasted. Servers are only busy 10--30% of the time on average, but they are often left on, while idle, utilizing 60% or more of peak power when in the idle state. We introduce a dynamic capacity management policy, AutoScale, that greatly reduces the number of servers needed in data centers driven by unpredictable, time-varying load, while meeting response time SLAs. AutoScale scales the data center capacity, adding or removing servers as needed. AutoScale has two key features: (i) it autonomically maintains just the right amount of spare capacity to handle bursts in the request rate; and (ii) it is robust not just to changes in the request rate of real-world traces, but also request size and server efficiency. We evaluate our dynamic capacity management approach via implementation on a 38-server multi-tier data center, serving a web site of the type seen in Facebook or Amazon, with a key-value store workload. We demonstrate that AutoScale vastly improves upon existing dynamic capacity management policies with respect to meeting SLAs and robustness.

How data center size impacts the effectiveness of dynamic power management

This paper presents a novel approach to correctly allocate resources in data centers, such that SLA violations and energy consumption are minimized. Our approach first analyzes historical workload traces to identify long-term patterns that establish a “base” workload. It then employs two techniques to dynamically allocate capacity: predictive provisioning handles the estimated base workload at coarse time scales (e.g., hours or days) and reactive provisioning handles any excess workload at finer time scales (e.g., minutes). The combination of predictive and reactive provisioning achieves a significant improvement in meeting SLAs, conserving energy, and reducing provisioning costs. We implement and evaluate our approach using traces from four production systems. The results show that our approach can provide up to 35% savings in power consumption and reduce SLA violations by as much as 21% compared to existing techniques, while avoiding frequent power cycling of servers.

Distributed, Robust Auto-Scaling Policies for Power Management in Compute Intensive Server Farms

While sleep states have existed for mobile devices and workstations for some time, these sleep states have not been incorporated into most of the servers in todays data centers. High setup times make data center administrators fearful of any form of dynamic power management, whereby servers are suspended or shut down when load drops. This general reluctance has stalled research into whether there might be some feasible sleep state (with sufficiently low setup overhead and/or sufficiently low power) that would actually be beneficial in data centers. This paper investigates the regime of sleep states that would be advantageous in data centers. We consider the benefits of sleep states across three orthogonal dimensions: (i) the variability in the workload trace, (ii) the type of dynamic power management policy employed, and (iii) the size of the data center. Our implementation results on a 24-server multi-tier testbed indicate that under many traces, sleep states greatly enhance dynamic power management. In fact, given the right sleep states, even a naive policy that simply tries to match capacity with demand, can be very effective. By contrast, we characterize certain types of traces for which even the “best” sleep state under consideration is ineffective. Our simulation results suggest that sleep states are even more beneficial for larger data centers.

Modeling the Impact of Workload on Cloud Resource Scaling

Power consumption accounts for a significant portion of a data centers operating expenses. Sadly, much of this power is wasted by servers that are left on even when there is no work to do. Dynamic power management aims to reduce power wastage in data centers by turning servers off when they are not needed. However, turning a server back on requires a setup time, which can adversely affect system performance. Thus, it is not obvious whether dynamic power management should be employed in a data center. In this paper, we analyze the effectiveness of dynamic power management in data centers under an M/M/k model via Matrix-analytic methods. We find that the effectiveness of even the simplest dynamic power management policy increases with the data center size, surpassing static power management when the number of servers exceeds 50, under realistic setup costs and server utilizations. Furthermore, we find that a small enhancement to traditional dynamic power management, involving delaying the time until a server turns off, can yield benefits over static power management even for data center sizes as small as 4 servers.

Exact analysis of the M/M/k/setup class of Markov chains via recursive renewal reward

Server farms today often over-provision resources to handle peak demand, resulting in an excessive waste of power. Ideally, server farm capacity should be dynamically adjusted based on the incoming demand. However, the unpredictable and time-varying nature of customer demands makes it very difficult to efficiently scale capacity in server farms. The problem is further exacerbated by the large setup time needed to increase capacity, which can adversely impact response times as well as utilize additional power.In this paper, we present the design and implementation of a class of Distributed and Robust Auto-Scaling policies (DRAS policies), for power management in compute intensive server farms. Results indicate that the DRAS policies dynamically adjust server farm capacity without requiring any prediction of the future load, or any feedback control. Implementation results on a 21 server test-bed show that the DRAS policies provide near-optimal response time while lowering power consumption by about 30% when compared to static provisioning policies that employ a fixed number of servers.

SOFTScale: stealing opportunistically for transient scaling

Cloud computing offers the flexibility to dynamically size the infrastructure in response to changes in workload demand. While both horizontal and vertical scaling of infrastructure is supported by major cloud providers, these scaling options differ significantly in terms of their cost, provisioning time, and their impact on workload performance. Importantly, the efficacy of horizontal and vertical scaling critically depends on the workload characteristics, such as the workloads parallelizability and its core scalability. In todays cloud systems, the scaling decision is left to the users, requiring them to fully understand the tradeoffs associated with the different scaling options. In this paper, we present our solution for optimizing the resource scaling of cloud deployments via implementation in OpenStack. The key component of our solution is the modelling engine that characterizes the workload and then quantitatively evaluates different scaling options for that workload. Our modelling engine leverages Amdahls Law to model service time scaling in scaleup environments and queueing-theoretic concepts to model performance scaling in scale-out environments. We further employ Kalman filtering to account for inaccuracies in the model-based methodology, and to dynamically track changes in the workload and cloud environment.

Autoscaling for Hadoop Clusters

The M/M/k/setup model, where there is a penalty for turning servers on, is common in data centers, call centers, and manufacturing systems. Setup costs take the form of a time delay, and sometimes there is additionally a power penalty, as in the case of data centers. While the M/M/1/setup was exactly analyzed in 1964, no exact analysis exists to date for the M/M/k/setup with