Wesley M. Felter

Energy management for commercial servers

Servers: high-end, multiprocessor systems running commercial workloads, have typically included extensive cooling systems and resided in custom-built rooms for high-power delivery. Recently, as transistor density and demand for computing resources have rapidly increased, even these high-end systems face energy-use constraints. Commercial-server energy management now focuses on conserving power in the memory and microprocessor subsystems. Because their workloads are typically structured as multiple application programs, system-wide approaches are more applicable to multiprocessor environments in commercial servers than techniques that primarily apply to single-application environments, such as those based on compiler optimizations.

A performance-conserving approach for reducing peak power consumption in server systems

The combination of increasing component power consumption, a desire for denser systems, and the required performance growth in the face of technology-scaling issues are posing enormous challenges for powering and cooling of server systems. The challenges are directly linked to the peak power consumption of servers.Our solution, Power Shifting, reduces the peak power consumption of servers minimizing the impact on performance. We reduce peak power consumption by using workload-guided dynamic allocation of power among components incorporating real-time performance feedback, activity-related power estimation techniques, and performance-sensitive activity-regulation mechanisms to enforce power budgets.We apply our techniques to a computer system with a single processor and memory. Power shifting adds a system power manager with a dynamic, global view of the systems power consumption to continuously re-budget the available power amongst the two components. Our contributions include:• Demonstration of the greater effectiveness of dynamic power allocation over static budgeting,• Evaluation of different power shifting policies,• Analysis of system and workload factors critical to successful power shifting, and• Proposal of performance-sensitive power budget enforcement mechanisms that ensure system reliability.

Practical DCB for Improved Data Center Networks

Storage area networking is driving commodity data center switches to support lossless Ethernet (DCB). Unfortunately, to enable DCB for all traffic on arbitrary network topologies, we must address several problems that can arise in lossless networks, e.g., large buffering delays, unfairness, head of line blocking, and deadlock. We propose TCP-Bolt, a TCP variant that not only addresses the first three problems but reduces flow completion times by as much as 70%. We also introduce a simple, practical deadlock-free routing scheme that eliminates deadlock while achieving aggregate network throughput within 15% of ECMP routing. This small compromise in potential routing capacity is well worth the gains in flow completion time. We note that our results on deadlock-free routing are also of independent interest to the storage area networking community. Further, as our hardware testbed illustrates, these gains are achievable today, without hardware changes to switches or NICs.

CPU packing for multiprocessor power reduction

Power and cooling considerations have moved to the forefront of modern system design. The restrictions placed upon systems by power and cooling requirements have focused much research on a variety of techniques to reduce maximum power and leakage. Simultaneously, efforts are being made to adapt microarchitectural features to the current needs of an application. We focus instead on adapting large scale resources to the current needs of a server farm. We study the efficacy of powering on and off CPUs in symmetric multiprocessors (SMP). We develop a number of different predictive and reactive techniques for identifying when cores should have their state altered. We present results for these policies and find a hybrid policy presents a reasonable balance between the time necessary to predict future needs and the accuracy of these predictions. It maintains 97% of the original system performance while reducing the energy per web interaction by 25%.

dCat: dynamic cache management for efficient, performance-sensitive infrastructure-as-a-service

In the modern multi-tenant cloud, resource sharing increases utilization but causes performance interference between tenants. More generally, performance isolation is also relevant in any multi-workload scenario involving shared resources. Last level cache (LLC) on processors is shared by all CPU cores in x86, thus the cloud tenants inevitably suffer from the cache flush by their noisy neighbors running on the same socket. Intel Cache Allocation Technology (CAT) provides a mechanism to assign cache ways to cores to enable cache isolation, but its static configuration can result in underutilized cache when a workload cannot benefit from its allocated cache capacity, and/or lead to sub-optimal performance for workloads that do not have enough assigned capacity to fit their working set. In this work, we propose a new dynamic cache management technology (dCat) to provide strong cache isolation with better performance. For each workload, we target a consistent, minimum performance bound irrespective of others on the socket and dependent only on its rightful share of the LLC capacity. In addition, when there is spare capacity on the socket, or when some workloads are not obtaining beneficial performance from their cache allocation, dCat dynamically reallocates cache space to cache-intensive workloads. We have implemented dCat in Linux on top of CAT to dynamically adjust cache mappings. dCat requires no modifications to applications so that it can be applied to all cloud workloads. Based on our evaluation, we see an average of 25% improvement over shared cache and 15.7% over static CAT for selected, memory intensive, SPEC CPU2006 workloads. For typical cloud workloads, with Redis we see 57.6% improvement (over shared LLC) and 26.6% improvement (over static partition) and with ElasticSearch we see 11.9% improvement over both.