Memory disaggregation has emerged as a strong alternative to traditional server systems for improved memory utilization and scalability. The compute nodes with a small local memory are connected to disaggregated memory pools through memory-semantic interconnects, such as CXL, that support high-bandwidth and low-latency memory access. A primary concern with these systems is high remote memory latency due to the presence of a network interconnect between CPU and memory. A hot page migration is generally deployed in a hybrid memory system to improve the locality of memory access. In this article, we identify the major challenges in the present OS-based or hardware-based techniques for page migration in disaggregated memory. To this end, we propose CosMoS , an architectural solution for Cos t-effective data Mo vement in a S calable disaggregated memory system that can predict, schedule, and optimize the movement of hot pages between local and remote memory. Our mechanism also eliminates long delays on the critical demand memory accesses to other remote pages that are obstructed during hot page movement. We evaluate CosMoS over many data-centric workloads. Our results show a 20% performance improvement with CosMoS compared to the state-of-the-art mechanism and an 86% improvement compared to the baseline for a large-scale disaggregated memory system.

Memory disaggregation has emerged as an alternative to traditional server architecture in data centers to target better memory utilization and higher scalability. It involves multiple independent compute nodes and remote memory pools that get hardware support through high-speed cache-coherent interconnects such as CXL. This paper introduces DRackSim, a simulation infrastructure for scalable disaggregated memory systems. DRackSim primarily models multiple compute nodes, memory pools, local/global memory managers, and a network interconnect for coherent memory access. An application-level simulation approach simulates an out-of-order x86 multi-core processor and a multi-level cache hierarchy at compute nodes. The network interface is simulated through a queue-based approach to handle remote memory access at multiple granularity. It also models a global memory manager for remote address space at the memory pools. Finally, we integrate a modified DRAMSim2 to perform local/remote memory simulation by declaring multiple instances of DRAMSim2. We rigorously validate DRackSim subsystems against Gem5 and a hardware prototype. Finally, we explore the design space by modeling various use-case scenarios for disaggregated memory systems and evaluate their performance over various HPC and cloud benchmarks.

Memory disaggregation is a solid alternative to traditional server systems that can overcome memory scalability issues in next-generation HPC data centers. In a rack-level disaggregated system, multiple compute nodes with small local memory rely on remote memory pools (memory nodes) to fulfill their memory demands. An in-network memory manager manages remote memory address space and allocates it to compute nodes which can access the memory at cache-line granularity using coherent interconnects such as CXL (or GenZ). However, the memory access cost is significantly increased due to the presence of the network. Even though a page migration system can exploit the locality of memory accesses, accessing a remote page starves the block-level requests. Further, page migrations introduce additional overheads which combined with starvation may even degrade the performance. All these issues require systematic evaluation of disaggregated memory systems to achieve improved designs. This paper presents a hardware mechanism for workload-aware data movement between compute and memory pools that significantly reduces the memory access cost. Firstly, our design enables centralized hot-page migration in a multi-tiered disaggregated memory that is aware of access patterns for individual compute nodes. Secondly, we analyze the complexities of accessing a remote memory page and propose a novel solution to eliminate starvation by serving all the remote memory requests at cache block granularity and by sharing bandwidth between page and block memory requests. Lastly, we add extra hardware support to get rid of additional overheads in a page migration system. We evaluate our designs over a variety of multi-threaded benchmarks using a cycle-level simulator which is specially designed to simulate a disaggregated memory system. Our design performs 10% to 100% better than traditional RDMA-based disaggregated systems that access remote memory at page granularity and 5% to 35% better than baseline disaggregated systems that use coherent interconnects for block-level access.

While anecdotally it is believed that Wi-Fi signal strength impacts the application performance, there is little work to quantity this connection. In this work, we explored a data-driven approach to model the relationship between Wi-Fi signal strength and the application performance in an indoor wireless environment. We discuss challenges in measurement in a noisy indoor Wi-Fi environment. Our experience and lessons learned are shared for the broader audience. While it was not possible to find definitive empirical models, our study shows that a trend in relationship can still be obtained. We conclude with lessons learned from this work.

Several network operations and management functions in the Internet depend on the traffic volume data represented as a traffic matrix (TM). Due to the difficulty in obtaining the TM data directly, several estimation techniques have been proposed in the literature. Most of the state-of-the-art techniques use subspace learning method that takes either one view of the data or multiple views gathered from different sources, to improve estimation accuracy. In this paper, we propose a multi-view subspace learning approach for accurate TM estimation using canonical correlation analysis. We define a TM view and show how multiple views of a TM, estimated with inexpensive techniques, can be used to estimate robust TMs. With experiments on Abilene network data, we show that the TMs estimated with the proposed multi-view learning technique have very low spatial and temporal error, compared to the other state-of-the-art techniques. We also show that the bias and variance in the estimated TMs are close to zero, which means that the estimated TMs can be very effective in capacity planning.

Dynamic adaptive streaming over HTTP (DASH) is widely used for video streaming on mobile devices. Ensuring a good quality of experience (QoE) for mobile video streaming is essential, as it severely impacts both the network and content providers’ revenue. Thus, a good rate adaptation algorithm at the client end that provides high QoE is critically important. Recently, a segment size-aware rate adaptation (SARA) algorithm was proposed for DASH clients. However, its performance on mobile clients has not been investigated so far. The main contributions of this article are twofold: (1) We discuss SARA’s implementation for mobile clients to improve the QoE in mobile video streaming, one that accurately predicts the download time for the next segment and makes an informed bitrate selection, and (2) we developed a new parametric QoE model to compute a cumulative score that helps in fair comparison of different adaptation algorithms. Based on our subjective and objective evaluation, we observed that SARA for mobile clients outperforms others by 17% on average, in terms of the Mean Opinion Score, while achieving, on average, a 76% improvement in terms of the interruption ratio. The score obtained from our new parametric QoE model also demonstrates that the SARA algorithm for mobile clients gives a better QoE among all the algorithms.

Use of technology in teaching-learning process has been increasing day-by-day. Chalkboard classroom is replaced by power point presentation in many places. Recently ICT is being used for the same. In this work, we proposed a system, Avabodhaka, where ICT is blended in traditional face-to-face classroom to improve teaching-learning process. The system along with its evaluation is presented in this article. It has been observed that students can learn better and more through this blended learning environment.

Recent advances in robotics and low-power embedded systems made three-dimensional (3D) mobile wireless sensor networks (MSNs) an effective solution for monitoring a field of interest (FoI). From a cost perspective, it is often important to ensure the desired coverage ratio for the FoI within a maximum allowable response (MAR) time, by using a minimum number of sensors in MSNs. The literature on determining the minimum number of sensors for the desired coverage ratio assumes that the FoI is unbounded to overcome the border effects. Since the entire sensing sphere of the sensors near the boundary may not be useful for the coverage, the number of sensors estimated without the border effects is lower than the actual value. In this paper, we estimate the minimum number of sensors required to achieve a desired coverage ratio in a given MAR time for a 3D FoI. We term this problem (α; V; T)-coverage problem, where a, V , and T are the desired coverage ratio, average speed of sensors, and MAR time, respectively. We assume straight line mobility model for the sensors and consider the border effects while deriving the expected sensing volume of a sensor useful in coverage. We also consider the restriction of sampling rate of the sensors in this analysis. We discuss the application of our analysis for a non-hyper-rectangle shaped FoI, random walk, and waypoint mobility models, and also the impact of neglecting the border effects. Our numerical and simulation results demonstrate the significance of border effects on the number of sensors and also the relationship between the coverage ratio, MAR time, sampling period, and the sensing range.

Many critical e-commerce and financial services are deployed on geo-distributed data centers for scalability and availability. Recent market surveys show that failure of a data center is inevitable resulting in a huge financial loss. Fault-tolerance in distributed data centers is typically handled by provisioning spare capacity to mask failure at a site. We argue that the operating cost and data replication cost (for data availability) must be considered in spare capacity provisioning along with minimizing the number of servers. Since the operating cost and client demand vary across space and time, we propose cost-aware capacity provisioning to minimize the total cost of ownership (TCO) for fault-tolerant data centers. We formulate the problem of spare capacity provisioning in fault-tolerant distributed data centers using mixed integer linear programming (MILP), with an objective of minimizing the TCO. The model accounts for heterogeneous client demand, data replication strategies (single and multiple site), variation in electricity price and carbon tax, and delay constraints while computing the spare capacity. Solving the MILP using real-world data, we observed a saving in the TCO to the tune of 35% compared to a model that minimizes the total number of servers and 43% compared to the model that minimizes the average response time. We demonstrate that our model is beneficial when the cost of electricity, carbon tax, and bandwidth vary significantly across the locations, which seems to be the problem for most of the operators.

Many popular e-commerce applications run on geo-distributed data centers requiring high availability. Fault-tolerant distributed data centers are designed by provisioning spare compute capacity to support the load of failed data center, apart from ensuring data durability. The main challenge during the planning phase is how to provision spare capacity such that the total cost of ownership (TCO) is minimized. While the literature handled spare capacity provisioning by minimizing the number of servers, variation in electricity cost and PUE corroborate the need to minimize the operating cost for capacity provisioning. We develop an MILP model for spare capacity provisioning for geo-distributed data centers with durability requirements. We consider spare capacity provisioning problem with the objective of minimizing TCO. We model variation in the demand, fluctuation in electricity prices across locations, cost of state replication, carbon tax across different countries, and delay constraints while formulating the optimization model. Solving the model shows that TCO is reduced while leveraging the electricity price variation and demand multiplexing. The proposed model outperforms the CDN model by 50% and the minimum server model by 34%. Results also demonstrate the effect of power usage effectiveness (PUE), latency, number of data centers and demand on the TCO.

Dynamic Adaptive Streaming over HTTP (DASH) enables the video player to adapt the bitrate of the video while streaming to ensure playback without interruptions even with varying throughput. A DASH server hosts multiple representations of the same video, each of which is broken down into small segments of fixed playback duration. The video bitrate adaptation is purely driven by the player at the endhost. Typically, the player employs an Adaptive Bitrate (ABR) algorithm, that determines the most appropriate representation for the next segment to be downloaded, based on the current network conditions and user preferences. The aim of an ABR algorithm is to dynamically manage the Quality of Experience (QoE) of the user during the playback. ABR algorithms manage the QoE by maximizing the bitrate while at the same time trying to minimize the other QoE metrics: playback start time, duration and number of buffering events, and the number of bitrate switching events. Typically, the ABR algorithms manage the QoE by using the measured network throughput and buffer occupancy to adapt the playback bitrate. However, due to the video encoding schemes employed, the sizes of the individual segments may vary significantly. For low bandwidth networks, fluctuation in the segment sizes results in inaccurate estimation the expected segment fetch times, thereby resulting in inaccurate estimation of the optimum bitrate. In this paper we demonstrate how the Segment-Aware Rate Adaptation (SARA) algorithm, that considers the measured throughput, buffer occupancy, and the variation in segment sizes helps in better management of the users' QoE in a DASH system. By comparing with a typical throughput-based and buffer-based adaptation algorithm under varying network conditions, we demonstrate that SARA manages the QoE better, especially in a low bandwidth network. We also developed AStream, an open-source Python-based emulated DASH-video player that was used to evaluate three different ABR algorithms and measure the QoE metrics with each of them.

Coverage and connectivity are important metrics used to evaluate the quality of service of wireless sensor networks (WSNs) monitoring a field of interest (FoI). Most of the literature assumes that the sensors are deployed directly in the FoI. In this paper we assume that the sensors are stochastically deployed outside the FoI. For such WSNs, we derive probabilistic expressions for k-coverage and connectivity using exact geometry. We validate our analysis and demonstrate its utility to estimate the minimum number of sensors required for a desired level of coverage and connectivity. We also demonstrate an on-campus traffic monitoring system to count the number of vehicles, detect the direction of vehicle, and to identify the vehicle (two-wheeler or four-wheeler) using sensors along both sides of the road.

Coverage is an important metric used to measure the quality of service of wireless sensor networks monitoring a field of interest (FoI). Existing literature on the coverage problem assumes that the sensors are deployed directly in the FoI. These results cannot be applied in some applications like canal water surface monitoring, because the sensors cannot be deployed on the water surface. In this paper, we analyze the coverage problem in applications where, the sensors are deployed uniformly at random outside the FoI near the boundary. We derive the expected value of the effective sensing area useful for k-coverage of the FoI using exact geometry. We demonstrate the utility of the analysis in estimation of the minimum number of sensors required for a desired level of coverage. With numerical results we show the impact of various parameters on the number of sensors.

Coverage is an important metric to measure the quality of service of a wireless sensor network monitoring a field of interest (FoI). From an energy perspective, it is often very important to maintain the desired coverage ratio with a minimum number of sensors. The literature on determining the critical sensor density (CSD) for the desired coverage ratio assumes that the FoI is unbounded or toroidal in shape. Although it is not a realistic assumption, it eliminates the border effects in analysis. Since the entire sensing area of the sensors near the boundary may not be useful for the coverage, the CSD estimated without the border effects is lower than the actual value. In this paper, we assume that the sensors are deployed uniformly at random in a convex polygon-shaped FoI and consider the border effects to derive the expected sensing area of a sensor used in the coverage. Next, we estimate the CSD required for the desired coverage ratio. We validate the analysis and demonstrate the impact of border effects on CSD using numerical results. Results show that our approach estimates the CSD better than another one that does not consider the exact geometry of the FoI.

A heterogeneous wireless sensor network (WSN) consists of sensors with unequal ranges of sensing and/or communication. In a dense WSN, a part of the region covered by a sensor may also be covered redundantly by a neighbouring sensor. In this paper, we analyse the redundancy in the coverage of a heterogeneous WSN and define the redundancy degree of a sensor. We follow a probabilistic approach to derive the expected redundancy degree of a sensor with a given number of sensors of each type in the neighbourhood. We demonstrate the accuracy of the analysis, and study the impact of the number of sensors of different types on the expected redundancy degree with numerical and simulation results. We also demonstrate an application of the redundancy analysis in the design of a heterogeneous WSN. We propose an algorithm to determine the minimum number of sensors of different types required to satisfy the desired coverage ratio and simultaneously minimise the cost of the network.

A new class of optical cross-connects (OXCs) is being proposed. These OXCs are highly scalable and possess the flexibility of 3-stage Clos networks so that various non-blocking properties, viz., rearrangeably nonblocking and strictly nonblocking (SNB) can be obtained from the same architecture. They are well-suited to implement SNB OXCs with high values of wavelength to fiber ratio, require lesser number of blocks than Clos network and support both fiber and wavelength scalability. The OXCs when implemented in three-dimensional micro-opto-electro-mechanical systems (3D MOEMS) technology provide greater advantages. A specific instance of the class of architectures is suitable for limited-range wavelength converters.

Wavelength converters in simple wavelength division multiplexed (WDM) networks helps in improving the network blocking performance by reducing the number of blocked demands arising out of the wavelength continuity constraint. However, cost and technological limitations make sparse wavelength conversion networks preferable. In this work, we propose a technique for wavelength converter placement under non-uniform traffic in the network. The algorithm that uses normalized cut based graph partitioning technique is extended for non-uniform traffic. A comparison of the performance under uniform and non-uniform traffic is given to highlight the effect of traffic on optimal placement. The algorithm is attractive due to lower complexity and efficiency.