Buffer management in shared-memory Time Warp systems

Fine-Grained Parallel and Distributed Spatial Stochastic Simulation of Biological Reactions

Hybrid memory composed of DRAM and persistent memory (PM) offers a promising way to realize large-capacity main memory supporting in-memory data storage and computing. However, traditional buffer management schemes focus on improving the hit ratio but lack awareness of the limitations of PM, e.g., slower write time and lower write endurance than DRAM. Therefore, developing new buffer management policies that can reduce costly write-backs of PM blocks while maintaining high performance for the hybrid buffer, is of paramount importance. Existing approaches mainly use a page-grained buffering policy, which will cause unnecessary data migrations between DRAM and PM, leading to a high number of disk I/Os and PM writes. Aiming to reduce I/O costs and PM writes, we propose a new buffer manager named HiBuffer for DRAM/PM-based hybrid memory systems. HiBuffer presents several novel ideas. First, it adopts multi-grained data layouts to manage the hybrid buffer cooperatively. In addition to the page granularity, we introduce Lines for the DRAM buffer and Sectors to the PM buffer, forming a buffer with three granularities, including Line, Sector, and Page. We prove that the multi-grained cooperative buffer management can deliver higher performance than existing page-grained schemes. Second, we propose a sector-grained method to migrate data from DRAM to PM, which can avoid unnecessary data movements and reduce PM writes. Third, we use an out-of-place updating mechanism to absorb updates in DRAM, which can further reduce the writes to PM. We compare HiBuffer with three existing schemes, including LRU, CLOCK-DWF, and MiniPage, on five synthetic workloads and the YCSB benchmark using real Intel Optane DC Persistent Memory. The results in terms of various metrics, including running time, PM writes, hit ratio, and disk I/Os, suggest the efficiency of HiBuffer. In particular, HiBuffer reduces the running time by up to 37.8% and the writes to PM by up to 83% compared to the competitors when evaluated on the YCSB benchmark.

We consider the management of FIFO buffers for network switches providing differentiated services. In each time step, an arbitrary number of packets arrive and only one packet can be sent. The buffer can store a limited number of packets and, due to the FIFO property, the sequence of sent packets has to be a subsequence of the arriving packets. The differentiated service model is abstracted by attributing each packet with a value according to its service level. A buffer management strategy can drop packets, and the goal is to maximize the sum of the values of sent packets.For only two different packet values, we introduce the account strategy and prove that this strategy achieves an optimal competitive ratio of \(\sqrt{2}-(\sqrt{5+4\sqrt{2}}-3)/2\approx 1.282\) if the buffer size tends to infinity and an optimal competitive ratio of \((\sqrt{13}-1)/2\approx 1.303\) for arbitrary buffer sizes. For general packet values, the simple preemptive greedy strategy (PG) is studied. We show that PG achieves a competitive ratio of \(\sqrt{3}\approx 1.732\) which is the best known upper bound on the competitive ratio of this problem. In addition, we give a lower bound of \(1+1/\sqrt{2}\approx 1.707\) on the competitive ratio of PG which improves the previously known lower bound. As a consequence, the competitive ratio of PG cannot be further improved significantly.

When a mobile node moves to a congested new BS with a general bottleneck link, a new buffer management scheme is proposed to improve utilization of the radio link. The proposed buffer management scheme determines the disposal of buffered packets depending on the degree of congestion. The proposed buffer management approach with considering the handover (Handover-based RED: HRED) provides a new disposal standard for the handover packets and improves throughput of existing flows. The simulation results confirm that the proposed buffer scheme can improve the link utilization performance about crowded wireless link.

As the power dissipation has become one of the critical design challenges in a sensor network environment, non- volatile memories such as STT-MRAM and flash memory will be used in the next generation sensor nodes. In this paper, we studied an efficient buffer management scheme considering the write limitation of STT-MRAM based main memory as well as the erase-before-write limitation of flash memory for storage device. The goal of proposed scheme is to minimize the number of write operations on STT- MRAM as well as the number of erase operations on flash memory. We showed through simulation that proposed scheme outperforms legacy buffer management schemes. Index Terms - Non-volatile memories, Low power, Buffer management, STT-MRAM, NAND flash memory.

It is widely believed that Time Warp is prone to two potential problems: an excessive amount of wasted, rolled back computation resulting from “rollback thrashing” behaviors, and inefficient use of memory, leading to poor performance of virtual memory and/or multiprocessor cache systems. An adaptive mechanism is proposed based on the Cancelback memory management protocol that dynamically controls the amount of memory used in the simulation in order to maximize performance. The proposed mechanism is adaptive in the sense that it monitors the execution of the Time Warp program, automatically adjusts the amount of memory used to reduce Time Warp overheads (fossil collection, Cancelback, the amount of rolled back computation, etc.) to a manageable level. The mechanism is based on a model that characterizes the behavior of Time Warp programs in terms of the flow of memory buffers among different buffer pools. We demonstrate that an implementation of the adaptive mechanism on a Kendall Square Research KSR-1 multiprocessor is effective in automatically maximizing performance while minimizing memory utilization of Time Warp programs, even for dynamically changing simulation models.

Router's buffer design and management strongly influence energy, area and performance of on-chip networks, hence it is crucial to encompass all of these aspects in the design process. At the same time, the NoC design cannot disregard preventing network-level and protocol-level deadlocks by devoting ad-hoc buffer resources to that purpose. In chip multiprocessor systems the coherence protocol usually requires different virtual networks (VNETs) to avoid deadlocks. Moreover, VNET utilization is highly unbalanced and there is no way to share buffers between them due to the need to isolate different traffic types. This paper proposes CUTBUF , a novel NoC router architecture to dynamically assign virtual channels (VCs) to VNETs depending on the actual VNETs load to significantly reduce the number of physical buffers in routers, thus saving area and power without decreasing NoC performance. Moreover, CUTBUF allows to reuse the same buffer for different traffic types while ensuring that the optimized NoC is deadlock-free both at network and protocol level. In this perspective, all the VCs are considered spare queues not statically assigned to a specific VNET and the coherence protocol only imposes a minimum number of queues to be implemented. Synthetic applications as well as real benchmarks have been used to validate CUTBUF , considering architectures ranging from 16 up to 48 cores. Moreover, a complete RTL router has been designed to explore area and power overheads. Results highlight how CUTBUF can reduce router buffers up to 33 percent with 2 percent of performance degradation, a 5 percent of operating frequency decrease and area and power saving up to 30.6 and 30.7 percent, respectively. Conversely, the flexibility of the proposed architecture improves by 23.8 percent the performance of the baseline NoC router when the same number of buffers is used.

End-user multi-flow services support is a crucial aspect of current and next generation mobile networks. This paper presents a dynamic buffer management strategy for HSDPA end-user multi-flow traffic with aggregated real-time and non-real-time flows. The scheme incorporates dynamic priority switching between the flows for transmission on the HSDPA radio channel. The end-to-end performance of the proposed strategy is investigated with an end-user multi-flow session of simultaneous VoIP and TCP-based downlink traffic using detailed HSDPA system-level simulations. Compared to an equivalent static buffer management scheme, the results show that end-to-end throughput performance gains in the non-real-time flow and better HSDPA channel utilization is attainable without compromising the real-time VoIP flow QoS constraints.

The present study is devoted to multihop Metropolitan Area Network architectures characterized by some buffer capacity at the nodes and routing algorithm slightly more complex than those already present in the literature. A new adaptive routing algorithm and a new buffer management scheme are presented, and their performance are compared to that of the best algorithms known in the literature. Both the new proposals exhibit significant improvements in throughput and delay with respect to the older ones, for various MAN topologies, sizes and buffer dimensions. These results, which in some cases are very close to the theoretical optimum, are achieved with an implementation complexity well within the limits of present day technology.

NAND flash storage devices are gaining popularity in consumer electronic devices, such as digital cameras, tablet PCs, and smartphones, due to their attractive features. However, write operations significantly limit the performance of the flash storage in consumer electronics devices due to their long latency and heavy overhead of garbage collection. Hence, NAND flash storage devices generally use RAM memory as a buffer to improve their write performance and to extend their lifetime. This paper proposes a novel buffer management algorithm (BMA), called as virtual-block-based buffer management scheme (VBBMS), which can make full use of both temporal and spatial localities at virtual-block level. The main difference with the existing BMAs is that the management granularity of VBBMS is a virtual block, while the management granularity of existing BMAs is a page or a block. In addition, VBBMS divides RAM into random request service region (RRSR), which is responsible for handling random requests, and sequential request service region (SRSR), which is responsible for handling sequential requests. When replacement happens, VBBMS selects victim virtual-block according to LRU policy in RRSR and first-in first-out (FIFO) policy in SRSR. VBBMS has been extensively evaluated under various realistic workloads, and our benchmark results show that VBBMS can improve buffer hit ratio, reduce response time, and erase times, compared to five existing BMAs.

Effective buffer management is crucial for efficient in-memory and streaming XQuery processing. We propose a buffer management scheme which combines static and dynamic analysis to keep main memory consumption low. Our approach relies on a technique that we call active garbage collection and which actively purges buffers at runtime based on the current status of query evaluation. We have built a prototype system for a practical fragment of XQuery which employs our buffer management scheme. The experimental results demonstrate the significant impact of combined static and dynamic analysis on reducing main memory consumption and running time.

Certain kinds of delay and disruption tolerant networks need the human support to send data between users without a contemporaneous end-to-end path between them. However, humans are socially selfish and tend to cooperate with whom they have a social relationship, fact that affects the overall performance. Memory space is a critical resource on mobile devices, and since the intermediate nodes have to store the messages, buffer management strategy becomes a crucial factor to achieve a satisfactory network performance. Very little attention has been directed towards applying the social characteristics of the users to mitigate the congestion in DTN. This paper presents a buffer management strategy for DTNs that takes into account a value associated with the social relationship strength among the users. We evaluated this policy in conjunction with the Epidemic and PRoPHET routing algorithms. By means of a thoroughly planned set of steady-state simulation experiments, we found that the proposed scheme can increase the delivery rate besides decreasing the average delay.

The distributed shared memory (DSM) architecture is considered to be one of the most likely parallel computing environment candidate for the near future because of its ease of system scalability and facilitation for parallel programming. However, a naive program based on shared memory execution on a DSM machine often deteriorates performance, because of the overhead involved for maintaining cache coherency particularly with frequent remote memory accesses. We show that careful buffer management of parallel join processing on DSM can produce considerable performance improvements in comparison with a naive implementation. We propose four buffer management strategies for parallel hash join processing on the DSM architecture and actually implement them on the HP Exemplar SPP 1600. The basic strategy is to begin with the hash join algorithm for the shared everything architecture and then to consider the memory locality of DSM by distributing the hash table and data pool buffers among the nodes. The results of four buffering strategies are analyzed in detail. Consequently, we can conclude that, in order to achieve high performance on a DSM machine, our buffer management strategy in which the memory access pattern is extracted and buffers are allocated in the local memory of nodes to minimize memory access cost is very efficient.

Switch and router architectures employing a shared buffer are known to provide high throughput, low delay, and high memory utilization. Superior performance of a shared-memory switch compared to switches employing other buffer strategies can be achieved by carefully implementing a buffer-management scheme. A buffer-sharing policy should allow all of the output interfaces to have fair and robust access to buffer resources. The sliding-window (SW) packet switch is a novel architecture that uses an array of parallel memory modules that are logically shared by all input and output lines to store and process data packets. The innovative aspects of the SW architecture are the approach to accomplishing parallel operation and the simplicity of the control functions. The implementation of a buffer-management scheme in a SW packet switch is dependent on how the buffer space is organized into output queues. This paper presents an efficient SW buffer-management scheme that regulates the sharing of the buffer space. We compare the proposed scheme with previous work under bursty traffic conditions. Also, we explain how the proposed buffer-management scheme can provide quality-of-service (QoS) to different traffic classes.

With the onset of multimedia applications, Quality of Service (QoS) in 802.11-based wireless ad hoc networks has been an essential area to look into. Standard wireless protocols such as IEEE 802.11b and IEEE 802.11e have been deemed unsuitable for multimedia traffic transmission in an ad hoc (multihop) network environment due to the unique network characteristics. As an ad hoc network is infrastructureless and distributed in nature, it gives rise to several challenges in QoS provisioning. In this thesis, we focus on the research of three key areas on QoS in ad hoc networks namely, buffer management in the medium access control (MAC) layer, cross-layer design and flow admission control. Firstly, we study MAC layer buffer management in the IEEE 802.11e protocol. The IEEE 802.11e MAC protocol provides QoS for multimedia traffic by probabilistic priority queueing. Ad hoc networks tend to generate a large volume of network control packets for route maintenance and these control packets are typically queued in the highest priority queue. As a result, higher priority control traffic is always being served, consuming most of the bandwidth and results in poor network performance for data traffic. We examine this problem and propose a dynamic buffer management scheme to mitigate it. By setting different threshold levels in different priority queues, we allow lower priority data traffic to be served when the higher priority control traffic is deemed to be excessive. Simulation of our proposed scheme has shown that our approach yielded better network performance compared to the standard protocol in terms of packet delivery ratio and average delay. Secondly, we study the cross-layer interaction between the MAC and Network layers to identify its impact on QoS. We select the IEEE 802.11e MAC protocol and Ad-hoc On Demand Distance Vector (AODV) routing protocol in our study since they are standardized protocols for wireless networks. In IEEE 802.11e MAC protocol, lower priority traffic has a higher probability of being blocked. Hence, the MAC layer will inform the Network layer of a broken link through the link notification function. In this case, AODV will re-route the paths that traverse this link. As a result, higher priority traffic which shares the same link as the lower priority traffic will undergo the same re-routing procedures. We termed this phenomenon path correlation effect. Under high traffic load, the path correlation effect will be more intense and hence degrade the QoS of higher priority traffic. To break the path correlation effect, we propose a cross-layer design called the Bundled Virtual Circuit (BVC). Our proposed BVC scheme not only addresses the path correlation issues, but also provides priority treatment from the Network layer to the MAC layer. Through our simulations, our proposed cross-layer design approach is proven to provide better network performance than the 802.11e/AODV protocols in terms of packet delivery ratio and delay. Finally, we examine flow admission control at the routing layer in ad hoc networks. Standard routing protocols such as AODV performs routing on a shortest path basis. In the presence of multiple routes, hotspots are likely to form among the intermediate nodes and thus lead to network congestion. In addition, existing routes are not well-protected from the introduction of new routes. We propose a flow admission control algorithm to be used with AODV. Our algorithm allows routes to be created through available bandwidth and congestion estimation. During the route discovery phase, each node will be gauged on its estimated available bandwidth and its backlog of packets in the MAC buffer. If a node does not have sufficient bandwidth, or has too many backlogged packets compared to the requested flow rate, it will not participate in the route discovery process. From simulations, our new scheme is able to provide better network performance than the standard AODV protocol in terms of delay. Our work in this thesis focuses on QoS provisioning in ad hoc networks from the MAC and network layer perspectives. By studying the network characteristics and protocols’ behavior, we investigate the various problems of providing QoS in an ad hoc network environment and propose three different schemes to improve the QoS provisioning at the MAC and network layers. We believe that our work is able to provide some insights to the research area on QoS in ad hoc networks.