CAFES: A framework for intrachip application modeling and communication architecture design

This paper describes CAFES, an extensible, open-source framework supporting several tasks related to high-level modeling and design of applications employing complex intrachip communication infrastructures. CAFES comprises several built-in models, including application, communication architecture, energy consumption and timing models. It also includes a set of generic and specific algorithms and additional supporting tools, which jointly with the cited models allow the designer to describe and evaluate applications requirements and constraints on specified communication architectures. Several examples of the use of CAFES underline the usefulness of the framework. Some of these are approached in this paper: (i) a realistic application captured at high-level that has its computation time estimated after mapping at the clock cycle level; (ii) a multi-application system that is automatically mapped to a large intrachip network with related tasks occupying contiguous areas in the chip layout; (iii) a set of mapping algorithms explored to define trade-offs between run time and energy savings for small to large intrachip communication architectures.     

Static timing analysis for modeling QoS in networks-on-chip

Networks-on-chip (NoCs) are used in a growing number of SoCs and multi-core processors. Because messages compete for the NoC’s shared resources, quality of service and resource allocation are major concerns for system designers. In particular, a model for the properties of packet delivery through the network is desirable. We present a methodology for packet-level static timing analysis in NoCs. Our methodology quickly and accurately gauges the performance parameters of a virtual-channel wormhole NoC without simulation. The network model can handle any topology, link capacities, and buffer sizes. It provides per-flow delay analysis that is orders-of-magnitude faster than simulation while being significantly more accurate than prior static modeling techniques. Using a carefully derived and reduced Markov chain, the model can statically represent the dynamic network state. Usage of the model in a placement optimization problem is shown as an example application.     

Self-stabilizing minimum degree spanning tree within one from the optimal degree

We propose a self-stabilizing algorithm for constructing a Minimum Degree Spanning Tree (MDST) in undirected networks. Starting from an arbitrary state, our algorithm is guaranteed to converge to a legitimate state describing a spanning tree whose maximum node degree is at most Δ^∗+1, where Δ^∗ is the minimum possible maximum degree of a spanning tree of the network.To the best of our knowledge, our algorithm is the first self-stabilizing solution for the construction of a minimum degree spanning tree in undirected graphs. The algorithm uses only local communications (nodes interact only with the neighbors at one hop distance). Moreover, the algorithm is designed to work in any asynchronous message passing network with reliable FIFO channels. Additionally, we use a fine grained atomicity model (i.e., the send/receive atomicity). The time complexity of our solution is O(mn²logn) where m is the number of edges and n is the number of nodes. The memory complexity is O(δlogn) in the send-receive atomicity model (δ is the maximal degree of the network).     

On-demand minimum cost benchmarking for intermediate dataset storage in scientific cloud workflow systems

Many scientific workflows are data intensive: large volumes of intermediate datasets are generated during their execution. Some valuable intermediate datasets need to be stored for sharing or reuse. Traditionally, they are selectively stored according to the system storage capacity, determined manually. As doing science on clouds has become popular nowadays, more intermediate datasets in scientific cloud workflows can be stored by different storage strategies based on a pay-as-you-go model. In this paper, we build an intermediate data dependency graph (IDG) from the data provenances in scientific workflows. With the IDG, deleted intermediate datasets can be regenerated, and as such we develop a novel algorithm that can find a minimum cost storage strategy for the intermediate datasets in scientific cloud workflow systems. The strategy achieves the best trade-off of computation cost and storage cost by automatically storing the most appropriate intermediate datasets in the cloud storage. This strategy can be utilised on demand as a minimum cost benchmark for all other intermediate dataset storage strategies in the cloud. We utilise Amazon clouds’ cost model and apply the algorithm to general random as well as specific astrophysics pulsar searching scientific workflows for evaluation. The results show that benchmarking effectively demonstrates the cost effectiveness over other representative storage strategies.     

On optimization of finite-difference time-domain (FDTD) computation on heterogeneous and GPU clusters

A model for the computational cost of the finite-difference time-domain (FDTD) method irrespective of implementation details or the application domain is given. The model is used to formalize the problem of optimal distribution of computational load to an arbitrary set of resources across a heterogeneous cluster. We show that the problem can be formulated as a minimax optimization problem and derive analytic lower bounds for the computational cost. The work provides insight into optimal design of FDTD parallel software. Our formulation of the load distribution problem takes simultaneously into account the computational and communication costs. We demonstrate that significant performance gains, as much as 75%, can be achieved by proper load distribution.     

Improving the yield of NoC-based systems through fault diagnosis and adaptive routing

We propose an effective and low cost method to increase the yield and the lifetime of torus NoCs. The method consists in detecting and diagnosing NoC interconnect faults using BIST structures and activating alternative paths for the faulty links. Alternative paths use the inherent redundancy of the torus topology, thus leading to minimal performance, area, and power overhead. We assume an extended interconnect fault model comprising stuck-at and pairwise shorts within a single link or between any two links in the network. Experimental results for a 3×3 NoC show that the proposed approach can correctly diagnose 93% of all possible interconnect faults and can mitigate 42% of those faults (representing 94.4% of the solvable faults) with a worst case performance penalty of 8% and 1% of area overhead. We also demonstrate the scalability of the method by presenting its application to larger NoCs.     

Online algorithms for advance resource reservations

We consider the problem of providing QoS guarantees to Grid users through advance reservation of resources. Advance reservation mechanisms provide the ability to allocate resources to users based on agreed-upon QoS requirements and increase the predictability of a Grid system, yet incorporating such mechanisms into current Grid environments has proven to be a challenging task due to the resulting resource fragmentation. We use concepts from computational geometry to present a framework for tackling the resource fragmentation, and for formulating a suite of scheduling strategies. We also develop efficient implementations of the scheduling algorithms that scale to large Grids. We conduct a comprehensive performance evaluation study using simulation, and we present numerical results to demonstrate that our strategies perform well across several metrics that reflect both user- and system-specific goals. Our main contribution is a timely, practical, and efficient solution to the problem of scheduling resources in emerging on-demand computing environments.     

An analytical model for multi-epidemic information dissemination

Contemporary distributed systems usually involve the spreading of information by means of ad-hoc dialogs between nodes (peers). This paradigm resembles the spreading of a virus in the biological perspective (epidemics). Such abstraction allows us to design and implement information dissemination schemes with increased efficiency. In addition, elementary information generated at a certain node can be further processed to obtain more specific, higher-level and more valuable information. Such information carries specific semantic value that can be further interpreted and exploited throughout the network. This is also reflected in the epidemical framework through the idea of virus transmutation which is a key component in our model. We establish an analytical framework for the study of a multi-epidemical information dissemination scheme in which diverse ‘transmuted epidemics’ are spread. We validate our analytical model through simulations. Key outcomes of this study include the assessment of the efficiency of the proposed scheme and the prediction of the characteristics of the spreading process (multi-epidemical prevalence and decay).     

Interest-Intended Piece Selection in BitTorrent-like peer-to-peer file sharing systems

BitTorrent is a popular peer-to-peer file sharing system and a target file shared through BitTorrent is partitioned into pieces and downloaded from multiple peers in parallel in order to shorten the download process. However, due to peer dynamics in P2P networks, rare pieces may be lost and thus lead to the so-called last piece problem. BitTorrent employs rarest-first piece selection algorithm to deal with this problem, but its efficacy is limited because each peer only has a local view of piece rareness. In this paper, we propose an Interest-Intended Piece Selection (IIPS) algorithm aiming at better alleviating the last piece problem while maintaining stable cooperation between peers. IIPS is named interest intended in that every IIPS peer favors pieces that, if downloaded, would increase the probability of being interesting to its cooperating peers. Simulation results show that IIPS achieves less occurrences of piece loss under tough conditions and slightly outperforms the BitTorrent’s rarest-first algorithm in terms of higher piece diversity.     

Throughput analysis and bandwidth allocation for IEEE 802.11 WLAN with hidden terminals

Motivated by observations from real world wireless local area network (WLAN) deployments, we develop in this paper a novel analytical model to characterize the saturation throughput of an IEEE 802.11-based access point (AP) and stations under the influence of hidden terminals. Unlike existing models, our model can accommodate different numbers of hidden nodes without increasing the model complexity. Given any number of hidden nodes, only four constraints are needed to describe the interaction between stations and the AP with the consideration of both uplink and downlink traffic. Simulation evaluation shows that our model predicts network performance accurately over a wide range of network sizes and indicates the existence of a throughput starvation problem. To address this problem, based on our model, we formulate a bandwidth allocation problem to optimize the network throughput and fairness under some predefined requirements by systematically tuning the AP and stations contention windows. Simulation results show that the starvation problem is resolved with our approach, and the target throughput is met.