Efficient data storage, a major concern in the modern computer industry, is mostly provided today by traditional magnetic disks. However, the cost of a disk transfer (measured in processor cycles) continues to increase with time, making disk accesses increasingly expensive.

Job migrations and network RAM are two major approaches for effectively using global memory resources in a workstation cluster, aimed at reducing page faults in each local workstation and improving the overall performance of cluster computing. Using either remote executions or preemptive migrations, a load sharing system is able to migrate a job from a workstation without sufficient memory space to a lightly loaded workstation with large idle memory space for the migrated job. In a network RAM system, if a job cannot find sufficient memory space for its working sets, it will utilize idle memory space from other workstations in the cluster through remote paging. Conducting tracedriven simulations, we have compared the performance and trade-offs of the two approaches and their impacts on job execution time and cluster scalability. Our study indicates that job-migration-based load sharing schemes are able to balance executions of jobs in a cluster well, while network RAM is able to satisfy data-intensive jobs which may not be migratable by sharing all the idle memory resources in a cluster. We also show that a network RAM cluster of workstations is scalable only if the network is sufficiently fast. Finally, we propose an improved load sharing scheme by combining job migrations with network RAM for cluster computing. This scheme uses remote execution to initially allocate a job to the most lightly loaded workstation and, if necessary, network RAM to provide a larger memory space for the job than would be available otherwise. The improved scheme has the merits of both job migrations and network RAM. Our experiments show its effectiveness and scalability for cluster computing.

Large scientific parallel applications demand large amounts of memory space. Current parallel computing platforms schedule jobs without fully knowing their memory requirements. This leads to uneven memory allocation in which some nodes are overloaded. This, in turn, leads to disk paging, which is extremely expensive in the context of scientific parallel computing. To solve this problem, we propose a new peer-to-peer solution called Parallel Network RAM. This approach avoids the use of disk and better utilizes available RAM resources. This approach will allow larger problems to be solved while reducing the computational, communication and synchronization overhead typically involved in parallel applications. 1.

Traditionally, operations with memory on other nodes (remote memory) in cluster environments interconnected with technologies like Gigabit Ethernet have been expensive with latencies several magnitudes slower than local memory accesses. Modern RDMA capable networks such as InfiniBand and Quadrics provide low latency of a few microseconds and high bandwidth of up to 10 Gbps. This has significantly reduced the latency gap between access to local memory and remote memory in modern clusters. Remote idle memory can be exploited to reduce the memory pressure on individual nodes. This is akin to adding an additional level in the memory hierarchy between local memory and the disk, with potentially dramatic performance improvements especially for memory intensive applications. In this paper, we take on the challenge to design a remote paging system for remote memory utilization in InfiniBand clusters. We present the design and implementation of a high performance networking block device (HPBD) over InfiniBand fabric, which serves as a swap device of kernel Virtual Memory (VM) system for efficient page transfer to/from remote memory servers. Our experiments show that using HPBD, quick sort performs only 1.45 times slower than local memory system, and up to 21 times faster than local disk. And our design is completely transparent to user applications. To the best of our knowledge, it is the first work of a remote pager design using InfiniBand for remote memory utilization.

So much has already been written about everything that you can't nd out anything about it. | James Thurber, Lanterns and Lances (1961) Loosely-coupled distributed systems haveevolved using message passing as the main paradigm for sharing information. Other paradigms used in loosely-coupled distributed systems, such as rpc, are usually implemented on top of an underlying message-passing system. On the other hand, in tightly-coupled architectures, such asmulti-processor machines, the paradigm is usually based on shared memory with its attractively simple programming model. The shared-memory paradigm has recently been extended for use in more loosely-coupled architectures and is known as distributed shared memory (dsm [153, 178,58]) in this context. This chapter discusses some of the issues involved in the design and implementation of such adsm in loosely-coupled distributed systems and brie y discusses related work in other elds. In dsm systems, processes share data transparently across node boundaries � data faulting, location, and movement are handled by thedsm system. Among other things, this allows parallel programs designed to use the shared-memory abstraction to execute without modi cation on a

Recent developments in the area of memory management have focused on reducing the eects of the disk latency problem. These developments include the use of a compressed cache and the utilisation of memory on remote hosts. Further developments in the areas of Quality of Service (QoS) provision and user-level virtual memory have provided the impetus for a more specialised run-time environment that is tailored for each individual application: an application-centric approach. Improvements on the disk latency problem have tended to be system-wide, i.e. all applications must participate in the scheme whether they would benet or not. A remote paging system incorporating QoS is presented here that oers individual applications the choice of participation or not. More specically, it allows them to have dierent paging schemes for dierent areas of virtual memory. The scheme also oers reliability and deterministic performance. 1 Introduction Recent developments in the area of memory managem...

The increasing use of high-bandwidth and low-latency networks make possible the use of remote (network) memory as an alternative to disk means of storing an application's data, because remote-to-local memory transfers over a modern interconnection network are faster than traditional disk-to-memory transfers. In this paper we explore the possibility of using the remote memory as (i) a (faster-than-disk) backing store, (ii) an extension of main memory accessed using single (remote) memory references, and (iii) as a combination of both. We use execution driven simulation to investigate the performance impact the use of remote memory has on several real programs. We conclude that even for today's low throughput networks, using remote memory as a place for storing (some) of an application's data may result in significant performance improvements, which will continue to get higher, as the disparity between disk transfer rates and network transfer rates continues to increase. 1 Introduction...

Transactions have been valued for their atomicity and recoverability properties that are useful to several systems, ranging from CAD environment to large-scale databases. Unfortunately, adding transaction support to an existing data repository was traditionally thought to be expensive, mostly due to the fact that the performance of transaction-based systems is usually limited by the performance of the magnetic disks that are used to hold the data repository. In this paper we describe how to use the collective main memory in a Network of Workstations (NOW) to improve the performance of transaction-based systems. We describe the design of our system and its implementation in two independent transaction-based systems, namely EXODUS, and RVM. We evaluate the performance of our prototype using several database benchmarks (like OO7 and TPC-A). Our experimental results indicate that our system delivers up to two orders of magnitude performance improvement compared to its predecessors. 1 Intro...

This paper explores the energy and delay issues that occur when some or all of the local storage is moved out of the embedded device, and into a remote network server. We demonstrate using the network to access remote storage in lieu of local DRAM results in significant power savings. Mobile applications continually demand additional memory, with traditional designs increasing DRAM to address this problem. Modern devices also incorporate low-power network links to support connected ubiquitous environments. Engineers then attempt to minimize utilization of the network due to its perceived large power consumption. This perception is misleading. For 1KB application "pages," network memory is more power efficient than one 2MB DRAM part when the mean time between page transfers exceeds 0.69s. During each transfer the application delay to the user is only 16ms.

Cluster applications that process large amounts of data, such as parallel scientific or multimedia applications, are likely to cause swapping on individual cluster nodes. These applications will perform better on clusters with network swapping support. Network swapping allows any cluster node with over-committed memory to use idle memory of a remote node as its backing store and to "swap" its pages over the network.