Y. Charlie Hu

Fine-grained power modeling for smartphones using system call tracing

Accurate, fine-grained online energy estimation and accounting of mobile devices such as smartphones is of critical importance to understanding and debugging the energy consumption of mobile applications. We observe that state-of-the-art, utilization-based power modeling correlates the (actual) utilization of a hardware component with its power state, and hence is insufficient in capturing several power behavior not directly related to the component utilization in modern smartphones. Such behavior arise due to various low level power optimizations programmed in the device drivers. We propose a new, system-call-based power modeling approach which gracefully encompasses both utilization-based and non-utilization-based power behavior. We present the detailed design of such a power modeling scheme and its implementation on Android and Windows Mobile. Our experimental results using a diverse set of applications confirm that the new model significantly improves the fine-grained as well as whole-application energy consumption accuracy. We further demonstrate fine-grained energy accounting enabled by such a fined-grained power model, via amanually implemented eprof, the energy counterpart of the classic gprof tool, for profiling application energy drain.

What is keeping my phone awake?: characterizing and detecting no-sleep energy bugs in smartphone apps

Despite their immense popularity in recent years, smartphones are and will remain severely limited by their battery life. Preserving this critical resource has driven smartphone OSes to undergo a paradigm shift in power management: by default every component, including the CPU, stays off or in an idle state, unless the app explicitly instructs the OS to keep it on! Such a policy encumbers app developers to explicitly juggle power control APIs exported by the OS to keep the components on, during their active use by the app and off otherwise. The resulting power-encumbered programming unavoidably gives rise to a new class of software energy bugs on smartphones called no-sleep bugs, which arise from mis-handling power control APIs by apps or the framework and result in significant and unexpected battery drainage. This paper makes the first advances towards understanding and automatically detecting software energy bugs on smartphones. It makes the following three contributions: (1) we present the first comprehensive study of real world no-sleep energy bug characteristics; (2) we propose the first automatic solution to detect these bugs based on the classic reaching definitions dataflow analysis algorithm; (3) we provide experimental data showing that our tool accurately detected all 17 known instances of no-sleep bugs and found 34 new bugs in the 73 apps examined.

The only constant is change: incorporating time-varying network reservations in data centers

In multi-tenant datacenters, jobs of different tenants compete for the shared datacenter network and can suffer poor performance and high cost from varying, unpredictable network performance. Recently, several virtual network abstractions have been proposed to provide explicit APIs for tenant jobs to specify and reserve virtual clusters (VC) with both explicit VMs and required network bandwidth between the VMs. However, all of the existing proposals reserve a fixed bandwidth throughout the entire execution of a job. In the paper, we first profile the traffic patterns of several popular cloud applications, and find that they generate substantial traffic during only 30%-60% of the entire execution, suggesting existing simple VC models waste precious networking resources. We then propose a fine-grained virtual network abstraction, Time-Interleaved Virtual Clusters (TIVC), that models the time-varying nature of the networking requirement of cloud applications. To demonstrate the effectiveness of TIVC, we develop Proteus, a system that implements the new abstraction. Using large-scale simulations of cloud application workloads and prototype implementation running actual cloud applications, we show the new abstraction significantly increases the utilization of the entire datacenter and reduces the cost to the tenants, compared to previous fixed-bandwidth abstractions.

Topology-aware routing in structured peer-to-peer overlay networks

Structured peer-to-peer (p2p) overlay networks like CAN, Chord, Pastry and Tapestry [19.14], [19.20], [19.17], [19.22] provide a self-organizing substrate for large-scale p2p applications. They can implement a scalable, fault-tolerant distributed hash table (DHT), in which any item can be located within a small number of routing hops using a small per-node routing table. These systems have been used in a variety of distributed applications, including distributed stores [19.7], [19.18], [19.10], [19.6], event notification, and content distribution [19.23], [19.5], [19.9], [19.4].

Bootstrapping energy debugging on smartphones: a first look at energy bugs in mobile devices

This paper argues that a new class of bugs faced by millions of smartphones, energy bugs or ebugs, have become increasingly prominent that already they have led to significant user frustrations. We take a first look at this emerging important technical challenge faced by the smartphones, ebugs, broadly defined as an error in the system (application, OS, hardware, firmware, external conditions or combination) that causes an unexpected amount of high energy consumption by the system as a whole. We first present a taxonomy of the kinds of ebugs based on mining over 39K posts (1.2M before filtering) from 4 online mobile user forum and mobile OS bug repositories. The taxonomy shows the highly diverse nature of smartphone ebugs. We then propose a roadmap towards developing a systematic diagnosing framework for debugging ebugs on smartphones.

Efficient network-coding-based opportunistic routing through cumulative coded acknowledgments

The use of random linear network coding (NC) has significantly simplified the design of opportunistic routing (OR) protocols by removing the need of coordination among forwarding nodes for avoiding duplicate transmissions. However, NC-based OR protocols face a new challenge: How many coded packets should each forwarder transmit? To avoid the overhead of feedback exchange, most practical existing NC-based OR protocols compute offline the expected number of transmissions for each forwarder using heuristics based on periodic measurements of the average link loss rates and the ETX metric. Although attractive due to their minimal coordination overhead, these approaches may suffer significant performance degradation in dynamic wireless environments with continuously changing levels of channel gains, interference, and background traffic. In this paper, we propose CCACK, a new efficient NC-based OR protocol. CCACK exploits a novel Cumulative Coded ACKnowledgment scheme that allows nodes to acknowledge network coded traffic to their upstream nodes in a simple way, oblivious to loss rates, and with practically zero overhead. In addition, the cumulative coded acknowledgment scheme in CCACK enables an efficient credit-based, rate control algorithm. Our evaluation shows that, compared to MORE, a state-of-the-art NC-based OR protocol, CCACK improves both throughput and fairness, by up to 20x and 124%, respectively, with average improvements of 45% and 8.8%, respectively.

Characterizing and modeling the impact of wireless signal strength on smartphone battery drain

Despite the tremendous market penetration of smartphones, their utility has been and will remain severely limited by their battery life. A major source of smartphone battery drain is accessing the Internet over cellular or WiFi connection when running various apps and services. Despite much anecdotal evidence of smartphone users experiencing quicker battery drain in poor signal strength, there has been limited understanding of how often smartphone users experience poor signal strength and the quantitative impact of poor signal strength on the phone battery drain. The answers to such questions are essential for diagnosing and improving cellular network services and smartphone battery life and help to build more accurate online power models for smartphones, which are building blocks for energy profiling and optimization of smartphone apps. In this paper, we conduct the first measurement and modeling study of the impact of wireless signal strength on smartphone energy consumption. Our study makes four contributions. First, through analyzing traces collected on 3785 smartphones for at least one month, we show that poor signal strength of both 3G and WiFi is routinely experienced by smartphone users, both spatially and temporally. Second, we quantify the extra energy consumption on data transfer induced by poor wireless signal strength. Third, we develop a new power model for WiFi and 3G that incorporates the signal strength factor and significantly improves the modeling accuracy over the previous state of the art. Finally, we perform what-if analysis to quantify the potential energy savings from opportunistically delaying network traffic by exploring the dynamics of signal strength experienced by users.

OpenMP for Networks of SMPs

In this paper, we present the first system that implements OpenMP on a network of shared-memory multiprocessors. This system enables the programmer to rely on a single, standard, shared-memory API for parallelization within a multiprocessor and between multiprocessors. It is implemented via a translator that converts OpenMP directives to appropriate calls to a modified version of the TreadMarks software distributed shared-memory (SDSM) system. In contrast to previous SDSM systems for SMPs, the modified TreadMarks system uses POSIX threads for parallelism within an SMP node. This approach greatly simplifies the changes required to the SDSM in order to exploit the intranode hardware shared memory. We present performance results for seven applications (Barnes-Hut, CLU, and Water from SPLASH-2, 3D-FFT from NAS, Red-Black SOR, TSP, and MGS) running on an SP2 with four four-processor SMP nodes. A comparison between the thread implementation and the original implementation of TreadMarks shows that using the hardware shared memory within an SMP node significantly reduces the amount of data and the number of messages transmitted between nodes and consequently achieves speedups that are up to 30% better than the original versions. We also compare SDSM against message passing. Overall, the speedups of multithreaded TreadMarks programs are within 7?30% of the MPI versions.

Hierarchical geographic multicast routing for wireless sensor networks

Wireless sensor networks comprise typically dense deployments of large networks of small wireless capable sensor devices. In such networks, multicast is a fundamental routing service for efficient data dissemination required for activities such as code updates, task assignment and targeted queries. In particular, efficient multicast for sensor networks is critical due to the limited energy availability in such networks. Multicast protocols that exploit location information available from GPS or localization algorithms are more efficient and robust than other stateful protocols as they avoid the difficulty of maintaining distributed state (multicast tree). Since localization is typically already required for sensing applications, this location information can simply be reused for optimizing multicast performance at no extra cost. Recently, two protocols were proposed to optimize two orthogonal aspects of location-based multicast protocols: GMR (Sanchez et al. GMR: Geographic multicast routing for wireless sensor networks. In Proceedings of the IEEE SECON, 2006) improves the forwarding efficiency by exploiting the wireless multicast advantage but it suffers from scalability issues when dealing with large sensor networks. On the other hand, HRPM (Das et al. Distributed hashing for scalable multicast in wireless ad hoc networks. IEEE TPDS 47(4):445–487, 2007) reduces the encoding overhead by constructing a hierarchy at virtually no maintenance cost via the use of geographic hashing but it is energy-inefficient due to inefficacies in forwarding data packets. In this paper, we present HGMR (hierarchical geographic multicast routing), a new location-based multicast protocol that seamlessly incorporates the key design concepts of GMR and HRPM and optimizes them for wireless sensor networks by providing both forwarding efficiency (energy efficiency) as well as scalability to large networks. Our simulation studies show that: (i) In an ideal environment, HGMR incurs a number of transmissions either very close to or lower than GMR, and, at the same time, an encoding overhead very close to HRPM, as the group size or the network size increases. (ii) In a realistic environment, HGMR, like HRPM, achieves a Packet Delivery Ratio (PDR) that is close to perfect and much higher than GMR. Further, HGMR has the lowest packet delivery latency among the three protocols, while incurring much fewer packet transmissions than HRPM. (iii) HGMR is equally efficient with both uniform and non-uniform group member distributions.

A Self-Organizing Flock of Condors

Condor provides high throughput computing by leveraging idle-cycles on off-the-shelf desktop machines. It also supports flocking, a mechanism for sharing resources among Condor pools. Since Condor pools distributed over a wide area can have dynamically changing availability and sharing preferences, the current flocking mechanism based on static configurations can limit the potential of sharing resources across Condor pools. This paper presents a technique for resource discovery in distributed Condor pools using peer-to-peer mechanisms that are self-organizing, fault-tolerant, scalable, and locality-aware. Locality-awareness guarantees that applications are not shipped across long distances when nearby resources are available. Measurements using a synthetic job trace show that self-organized flocking reduces the maximum job wait time in queue for a heavily loaded pool by a factor of 10 compared to without flocking. Simulations of 1000 Condor pools are also presented and the results confirm that our technique discovers and utilizes nearby resources in the physical network.