Anjur Sundaresan Krishnakumar

Supporting soft real-time tasks in the xen hypervisor

Virtualization technology enables server consolidation and has given an impetus to low-cost green data centers. However, current hypervisors do not provide adequate support for real-time applications, and this has limited the adoption of virtualization in some domains. Soft real-time applications, such as media-based ones, are impeded by components of virtualization including low-performance virtualization I/O, increased scheduling latency, and shared-cache contention. The virtual machine scheduler is central to all these issues. The goal in this paper is to adapt the virtual machine scheduler to be more soft-real-time friendly. We improve two aspects of the VMM scheduler -- managing scheduling latency as a first-class resource and managing shared caches. We use enterprise IP telephony as an illustrative soft real-time workload and design a scheduler S that incorporates the knowledge of soft real-time applications in all aspects of the scheduler to support responsiveness. For this we first define a laxity value that can be interpreted as the target scheduling latency that the workload desires. The load balancer is also designed to minimize the latency for real-time tasks. For cache management, we take cache-affinity into account for real time tasks and load-balance accordingly to prevent cache thrashing. We measured cache misses and demonstrated that cache management is essential for soft real time tasks. Although our scheduler S employs a different design philosophy, interestingly enough it can be implemented with simple modifications to the Xen hypervisors credit scheduler. Our experiments demonstrate that the Xen scheduler with our modifications can support soft real-time guests well, without penalizing non-real-time domains.

On the accuracy of signal strength-based estimation techniques

In this paper, we address the problem of finding the inherent uncertainty of signal strength-based location estimation techniques. We propose a mathematical model for mapping uncertainty in signal strength space to uncertainty in physical space. We then analyze this model to compute the minimum value of the uncertainty in location estimation using signal strength measurements. The results of this analysis are used to draw conclusions about the dependence of the minimum uncertainty of various factors such as the signal variance, number of APs, distance between the APs and the propagation constant. We provide an argument linking the minimum uncertainty with a lower limit on the median error in location estimation using classification techniques.

The theory and practice of signal strength-based location estimation

Location estimation enables many innovative applications and is an exciting area of research. With the growing use of wireless technology in enterprise networks, it is an interesting technical challenge to develop techniques for indoor location estimation that leverage the deployed wireless infrastructure. Elegant, cost-effective techniques would present a compelling business proposition. In this paper, we summarize various approaches researchers have studied for the problem of indoor location estimation, concentrating on signal-strength based techniques directed towards 802.11 wireless networks. We also summarize an interesting insight into the best possible accuracy achievable by any technique due to limits imposed by the physical behavior of the radio signal. We identify that the issues driving research work in this area will not only be location accuracy, but other factors like deployment ease, management simplicity, adaptability, and cost of ownership and maintenance. We conclude with some open research problems in this area

Performance implications of hosting enterprise telephony applications on virtualized multi-core platforms

Virtualization technology has gained significant adoption in various domains as a means to lower costs and enable greener solutions. Recently, there has been a significant amount of interest in employing virtualization technology in the telecommunications domain in order to save costs through server consolidation and to provide energy-efficient solutions. The availability of high-end multi-core servers provides powerful platforms for deployment. However, the telecommunications domain poses unique challenges for virtualization technology to be successfully deployed even in these compute-rich multi-core environments. This work discusses these challenges. It provides a detailed analysis of the performance implications of hosting enterprise IP telephony infrastructure in virtualized environments. Unlike signaling applications that are comparatively more tolerant of underlying platform performance, media applications are far more demanding. Our work, therefore, focuses on the performance of media applications (media server, voice-mail, etc.) in virtualized environments. We develop a model for workloads used in enterprise IP telephony. We then evaluate the impact of various hypervisor scheduler and I/O parameters in order to determine good parameter settings for such workloads. Our experiments use the Xen virtualization platform. The results presented in this work will be useful for telecommunication solution providers to understand the capabilities and limitations of virtual environments when deploying their applications.

Location Estimation in Wireless Networks: A Bayesian Approach

We present a Bayesian hierarchical model for indoor location estimation in wireless networks. We demonstrate that our model achieves accuracy that is similar to other published models and algorithms. By harnessing prior knowledge, our model drastically reduces the requirement for training data as compared with existing approaches.

Infrastructure-based location estimation in WLAN

This paper focuses on fundamental system deployment aspects of location estimation in 802.11-based wireless networks. We concentrate on adaptable infrastructure-based approaches, where sniffers measure received signal strength from clients to locate them. Our implementation experience and experimental results show that sniffer-based location estimation is feasible and works well provided some important rules are followed. By studying data over a 6-month period, we observe that adaptation of models is necessary for good location estimation, and we show that our techniques enable location estimation with minimal profiling. We also present an intriguing client-assisted approach for location estimation where a client, APs and sniffers collaborate to locale a terminal in enterprise (infrastructure-mode) wireless networks.

Estimating signal strength coverage for a wireless access point

Estimating the signal strength coverage of a wireless access point (AP), given its desired location at a site, is essential in many instances, e.g., to determine optimal AP placement. In this paper, we present a new non-parametric modeling solution, based on sampling, for this problem. We use sample measurements from several fixed APs in a novel way to compute a signal coverage model for an AP at a random location. Our technique automatically takes into account the signal strength propagation characteristics of the site and also allows for an efficient deployment methodology. We present an experimental study that validates the practical usability of the proposed technique.

Hypervisor-assisted application checkpointing in virtualized environments

There are two broad categories of approaches used for checkpointing: application-transparent and application-assisted. Typically, application-assisted approaches provide a more flexible and light-weight mechanism but require changes to the application. Although most applications run well under virtualization (e.g. Xen which is being adopted widely), the addition of application-assisted checkpointing - used for high availability - causes performance problems. This is due to the overhead of key system calls used by the checkpointing techniques under virtualization. To overcome this, we introduce the notion of hypervisor-assisted application checkpointing with no changes to the guest operating system. We present the design and a Xen-based implementation of our family of application checkpointing techniques. Our experiments show performance improvements of 4× to 13× in the primitives used for supporting high availability compared to purely user-level approaches.

Measuring video quality degradation using face detection

Ensuring end-to-end video quality requires monitoring quality in real-time (in-service) and taking counter-measures in times of adverse network conditions. Such application-layer QoS assurance mechanisms require light-weight video quality metrics that can be implemented with low computational and communication overheads. In this paper, we propose a novel video quality metric for video conferencing-type applications that accurately reflects user opinion and is light-weight for realtime operations. Our motivation is to exploit the characteristics of the video content in such applications, i.e. few speakers with limited motion. Our metric, Simplified Perceptual Quality Region (SPQR), relies on detecting the location of a speakers face in sent and received video frames and comparing the locations between the corresponding frames in the two streams to identify discrepancies as a sign of video quality degradation. Our experiments show that face locations can be determined in realtime by sampling few frames every second. SPQR is a reduced-reference metric that requires minimal transmission overhead between the sender and receiver through a separate channel to communicate the reduced features. In this paper, we present an empirical evaluation of the performance of SPQR using a video phone application. We first show that SPQR effectively detects video quality degradation. Second, we compare our proposed metric to two well-accepted full-reference techniques appropriate for offline analysis, namely PSNR and VQM, and show that SPQR tracks both metrics well. Finally, we show that low grade sampling yields SPQR values comparable to PSNR and VQM scores and thus enabling a light-weight implementation.

Seamless session mobility using short range communication technologies

Rapid adoption of smartphones and tablets in recent years has changed the way people communicate with each other and operate in the work environment. The use of personal devices in the enterprise has become commonplace. These devices have given workers the flexibility of being mobile and still participate in collaboration sessions. However, transferring sessions between devices is not very seamless and causes disruptions in the sessions, resulting in an unsatisfactory user experience. This work presents an architecture and implementation for seamless session mobility across different types of devices and hardware/software platforms. This architecture is not restricted to a specific collaboration server and can accommodate multi-vendor environments.