WattScope is a system for non-intrusively estimating the power consumption of individual applications using external measurements of a server's aggregate power usage and without requiring direct access to the server's operating system or applications. Our key insight is that, based on an analysis of production traces, the power characteristics of datacenter workloads, e.g., low variability, low magnitude, and high periodicity, are highly amenable to disaggregation of a server's total power consumption into application-specific values. WattScope adapts and extends a machine learning-based technique for disaggregating building power and applies it to server- and rack-level power measurements that are already available in datacenters. We evaluate WattScope's accuracy on a production workload and show that it yields high accuracy, e.g., often <∼10% normalized mean absolute error, and is thus a potentially useful tool for datacenters in externally monitoring application-level power usage.
This content will become publicly available on November 1, 2024
WattScope: Non-intrusive application-level power disaggregation in datacenters
Datacenter capacity is growing exponentially to satisfy the increasing demand for many emerging computationally-intensive applications, such as deep learning. This trend has led to concerns over datacenters’ increasing energy consumption and carbon footprint. The most basic prerequisite for optimizing a datacenter’s energy- and carbon-efficiency is accurately monitoring and attributing energy consumption to specific users and applications. Since datacenter servers tend to be multi-tenant, i.e., they host many applications, server- and rack-level power monitoring alone does not provide insight into the energy usage and carbon emissions of their resident applications. At the same time, current application-level energy monitoring and attribution techniques are intrusive: they require privileged access to servers and necessitate coordinated support in hardware and software, neither of which is always possible in cloud environments. To address the problem, we design WattScope, a system for non-intrusively estimating the power consumption of individual applications using external measurements of a server’s aggregate power usage and without requiring direct access to the server’s operating system or applications. Our key insight is that, based on an analysis of production traces, the power characteristics of datacenter workloads, e.g., low variability, low magnitude, and high periodicity, are highly amenable to disaggregation of a server’s total power consumption into application-specific values. WattScope adapts and extends a machine learning-based technique for disaggregating building power and applies it to server- and rack-level power meter measurements that are already available in data centers. We evaluate WattScope’s accuracy on a production workload and show that it yields high accuracy, e.g., often 10% normalized mean absolute error, and is thus a potentially useful tool for datacenters in externally monitoring application-level power usage.
more »
« less
- Award ID(s):
- 2105494
- NSF-PAR ID:
- 10496891
- Publisher / Repository:
- Performance Evaluation
- Date Published:
- Journal Name:
- Performance Evaluation
- Volume:
- 162
- Issue:
- C
- ISSN:
- 0166-5316
- Page Range / eLocation ID:
- 102369
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
more » « less
-
null (Ed.)Low-latency online services have strict Service Level Objectives (SLOs) that require datacenter systems to support high throughput at microsecond-scale tail latency. Dataplane operating systems have been designed to scale up multi-core servers with minimal overhead for such SLOs. However, as application demands continue to increase, scaling up is not enough, and serving larger demands requires these systems to scale out to multiple servers in a rack. We present RackSched, the first rack-level microsecond-scale scheduler that provides the abstraction of a rack-scale computer (i.e., a huge server with hundreds to thousands of cores) to an external service with network-system co-design. The core of RackSched is a two-layer scheduling framework that integrates inter-server scheduling in the top-of-rack (ToR) switch with intra-server scheduling in each server. We use a combination of analytical results and simulations to show that it provides near-optimal performance as centralized scheduling policies, and is robust for both low-dispersion and high-dispersion workloads. We design a custom switch data plane for the inter-server scheduler, which realizes power-of-k- choices, ensures request affinity, and tracks server loads accurately and efficiently. We implement a RackSched prototype on a cluster of commodity servers connected by a Barefoot Tofino switch. End-to-end experiments on a twelve-server testbed show that RackSched improves the throughput by up to 1.44x, and scales out the throughput near linearly, while maintaining the same tail latency as one server until the system is saturated.more » « less
-
Modern Information Technology (IT) servers are typically assumed to operate in quiescent conditions with almost zero static pressure differentials between inlet and exhaust. However, when operating in a data center containment system the IT equipment thermal status is a strong function of the non- homogenous environment of the air space, IT utilization workloads and the overall facility cooling system design. To implement a dynamic and interfaced cooling solution, the interdependencies of variabilities between the chassis, rack and room level must be determined. In this paper, the effect of positive as well as negative static pressure differential between inlet and outlet of servers on thermal performance, fan control schemes, the direction of air flow through the servers as well as fan energy consumption within a server is observed at the chassis level. In this study, a web server with internal air-flow paths segregated into two separate streams, each having dedicated fan/group of fans within the chassis, is operated over a range of static pressure differential across the server. Experiments were conducted to observe the steady-state temperatures of CPUs and fan power consumption. Furthermore, the server fan speed control scheme’s transient response to a typical peak in IT computational workload while operating at negative pressure differentials across the server is reported. The effects of the internal air flow paths within the chassis is studied through experimental testing and simulations for flow visualization. The results indicate that at higher positive differential pressures across the server, increasing server fans speeds will have minimal impact on the cooling of the system. On the contrary, at lower, negative differential pressure server fan power becomes strongly dependent on operating pressure differential. More importantly, it is shown that an imbalance of flow impedances in internal airflow paths and fan control logic can onset recirculation of exhaust air within the server. For accurate prediction of airflow in cases where negative pressure differential exists, this study proposes an extended fan performance curve instead of a regular fan performance curve to be applied as a fan boundary condition for Computational Fluid Dynamics simulations.more » « less
-
As a result of constantly increasing data center utilization, many challenges have appeared for thermal engineers over the last few decades. Advanced cooling systems for servers are of significant interest, particularly, technologies which can also reduce electricity usage. One known technology called Organic Rankine Cycle (ORC) is considered a viable alternative for this purpose. It can both the heat from a server and then transfer the server heat into a power cycle to generate electricity. This study consists of the design and construction of an experimental prototype of 20kW of waste heat, representing two common rack servers operating at full capacity. The range of server waste heat temperatures is between 60°C to 85°C, which is far below the normal operating range for ORCs. This ultra-low-grade waste heat leads to an expected thermal efficiency between 2%-8%. Tests on the experimental rig showed a maximum thermal efficiency of 3.33%. The system is both absorbing all the waste heat from the data center and at the same time providing an economic benefit back to the data center in form of electricity. Through experimental investigation, this study provides the first evidence for using ORC system as a valid solution for ultra-low-grade waste heat recovery.more » « less
-
Empirical performance measurements of computer systems almost always exhibit variability and anomalies. Run-to-run and server-to-server variations are common for CPU, memory, disk, and network performance characteristics. In our previous work, we focused on taming performance variability for memory, disk, and network and established an interactive analysis service at: https://confirm.fyi/ to help users of the CloudLab testbed better plan and conduct their experiments. In this paper, we describe our analysis of CPU variability based on over 1.3M performance measurements from nearly 1,800 servers and present our initial findings. The focus of this work is on capturing hardware variability, which can make repeatable experiments more difficult and can impact conclusions; it it this important for systems researchers to understand. (We note that, though we do not study it in this work, in the cloud, multi-tenancy and resource sharing an exacerbate the problem.) Variability also inevitably impacts performance and operation of middleware and high-level applications, contributing to the straggler problems in many domains, including HPC, Big Data, and Machine Learning, and on many types of cyberinfrastructures. We analyze the data from the CloudLab servers allocated in an exclusive fashion, with no virtualization. While our analysis focuses on the testbed that aims to promote reproducible research, we believe our approach and the findings can be of value to people who manage, analyze, and utilize shared computing resources in supercomputers, clouds, and datacenters.more » « less
