Supporting Huge Address Spaces in a Virtual Machine for Java on a Cluster

Remote Direct Memory Access (RDMA) is a technology to update a remote machine's memory without intervention at the receiver side. We evaluate where RDMA can be usefully applied and where it is a loss in Object-Oriented DSM systems. RDMA is difficult to use in modern OO-DSMs due to their support for large address spaces, advanced protocols, and heterogeneity. First, a communication pattern that is based on objects reduces the applicability of bulk RDMA. Second, large address spaces (meaning far larger than that of a single machine) and large numbers of machines require an address space translation scheme to map an object at different addresses on different machines. Finally, RDMA usage is hard since without polling (which would require source code modifications), incoming RDMA messages are hard to notice on time. Our results show that even with RDMA, update protocols are slower than invalidation protocols. But RDMA can be successfully applied to fetching of objects in an invalidation protocol and improves performance by 20.6%.

With the growing importance of in-memory data processing, cloud service providers have launched large memory virtual machine services to accommodate memory intensive workloads. Such large memory services using low volume scaled-up machines are far less cost-efficient than scaled-out services consisting of high volume commodity servers. By exploiting memory usage imbalance across cloud nodes, disaggregated memory can scale up the memory capacity for a virtual machine in a cost-effective way. Disaggregated memory allows available memory in remote nodes to be used for the virtual machine requiring more memory than its locally available memory. It supports high performance with the faster direct memory while satisfying the memory capacity demand with the slower remote memory. This paper proposes a new hypervisor-integrated disaggregated memory system for cloud computing. The hypervisor-integrated design has several new contributions in its disaggregated memory design and implementation. First, with the tight hypervisor integration, it investigates a new page management mechanism and policy tuned for disaggregated memory in virtualized systems. Second, it restructures the memory management procedures and relieves the scalability concern for supporting large virtual machines. Third, exploiting page access records available to the hypervisor, it supports application-aware elastic block sizes for fetching remote memory pages with different granularities. Depending on the degrees of spatial locality for different regions of memory in a virtual machine, the optimal block size for each memory region is dynamically selected. The experimental results with the implementation integrated to the KVM hypervisor, show that the disaggregated memory can provide on average 6 percent performance degradation compared to the ideal local-memory only machine, even though the direct memory capacity is only 50 percent of the total memory footprint.

Heterogeneous Network-Based Concurrent Computing Systems

SummaryScience Gateways provide scientists with tools for creating, executing, and monitoring scientific experiments on multiple resource infrastructures. Apache Airavata abstracts interactions between gateways and distributed computing infrastructures like Extreme Science and Engineering Discovery Environment, international grids, and campus clusters. Airavata consists of several component services such as the API server, Orchestrator, Workflow Interpreter, Credential Store, and Application Factory. In addition, Airavata uses third party software, including RabbbitMQ for messaging, MySQL for production database management, and Apache Zookeeper for internal communications. In this paper, we discuss our initial experiences with leveraging open source technologies to manage Airavata and its dependent components to deploy, detect, and restart failed components in an auto‐scaling platform. Such capabilities will allow Airavata services to be deployed in a wide area, large Virtual Machine (VM) based cluster, and a developer's laptop. The emerging technologies in cloud computing and Big Data that address these needs are the following: Docker, Marathon, and Apache Mesos. Docker is a Linux‐based lightweight container that allows different applications to run isolated from each other but safely share the machine's resources. Docker images of applications can be published in registries and retrieved for execution in the target infrastructures. Marathon provides a cluster‐wide init and control system for services, including Docker containers. Mesos provides a cluster‐wide framework to schedule tasks based on fine‐grained resource needs. Mesosphere provides the packages, scripts, and web interface to ease the use of these technologies. We present the design, experience, and lessons learned from integrating Mesos, Docker, and Marathon with Apache Airavata. Copyright © 2015 John Wiley & Sons, Ltd.

Consolidating multiple virtual machines into a single server has been widely adopted in cloud computing to improve system utilization. However, the sharing of physical CPUs among virtual machines in consolidated systems poses a new challenge in providing an illusion of continuous CPU execution to the guest operating systems (OS). Due to the time-sharing of physical CPUs, the execution of a guest OS is periodically interrupted, while the guest OS may not be aware of the discontinuity of virtual time against the real time. The virtual time discontinuity problem causes the delayed processing of critical OS operations, such as interrupt handling and lock processing. Although there have been several prior studies to mitigate the problem, they address only a subset of symptoms, require the modification of guest OSes, or change the processor architecture. This paper proposes a novel way to comprehensively reduce the inefficiency of guest OS execution in consolidated systems. It migrates the short-lived critical OS tasks to dedicated micro-sliced cores, minimizing the delays caused by time sharing. The hypervisor identifies the critical OS tasks without any OS intervention, and schedules the critical code sections onto the dynamically partitioned cores at runtime. The dedicated micro-sliced cores employ a short sub-millisecond quantum to minimize the response latencies for consolidated virtual machines. By readily servicing the critical tasks, the proposed scheme can minimize the adverse artifact of virtual machine consolidation without any modification of guest OSes.

Since a virtual machine independently uses its own caching policy, redundant disk operations exacerbate the I/O virtualization overhead when virtual machines access large amounts of data on shared storage. This paper presents XHive, an efficient cooperative caching system that is implemented at the virtualization layer, for consolidated environments. Our proposed scheme globally manages buffer caches of consolidated virtual machines in order to accommodate a shared working set in machine memory. A singlet, which is a block cached solely by a virtual machine, is preferentially given more chances to be cached in machine memory by XHive, when it is evicted by a guest operating system. For efficient use of limited memory, singlets are cached in memory that is collaboratively donated from idle memory of virtual machines. Our evaluation shows that XHive significantly reduces disk I/O operations for shared working sets, thereby achieving high read performance and scalability. Improved scalability enables a high degree of workload consolidation with respect to virtual machines that have shared working sets.

Some aspects of research into distributed shared memory (DSM) systems which concern reliability are presented. The research is concerned with applications that use shared memory in single-site systems, and their extension to operation in a distributed environment. A DSM system consists of a shared address space that is accessed by logical operations. In the author's model processes create shared memory by using a segment; the segment's size, name, and access protection are defined when the memory is created. Processes locate and attach the segment into their virtual memory address space by name. When a process is finished with the segment, it may be detached. The last detach of a segment destroys the data. The shared portion of a process's address space is partitioned into data pages, which are replicated. System implementation, distributed architecture, and issues affecting DSM reliability are discussed. >

It is well known that contention is one of the factors that limit the performance of high performance parallel computing systems that implement distributed shared memory (DSM). On cache-coherent DSM multiprocessors, one of the most intrusive forms of contention is the one that occurs at the network interface of a node, due to simultaneous requests for remote memory accesses. This thesis focuses on the issue of resolving remote memory access contention on hardware DSM systems, and on the performance impact of implementing contention resolution algorithms. The Simultaneous Optical Multiprocessor Exchange Bus (SOME-Bus) is a low-latency, high-bandwidth, fiber-optic interconnection network supporting DSM with a unique feature that every processor is connected to the other processor through a dedicated output channel. The thesis presents simple but effective contention resolution algorithms that rely on the information of average channel utilization, average channel waiting time, number of messages in the channel queue or short-term average channel waiting time reported by each node, and uses simulation to examine the performance of the algorithms on the SOMEBus multiprocessor. The algorithms detect potential hot spots and resolve contention using dynamic block migration, which is a novel approach for migrating home nodes on a block-level basis and balancing remote memory accesses across the nodes of the system. The algorithms require minimal support from the operating system and all the necessary message transactions happen on the hardware level. Experiments with eight parallel codes on a 64-processor SOME-Bus show that algorithms yield significant performance improvements such as reduction in the execution times, number of remote memory accesses, average channel waiting times and average network latencies. Modified contention resolution algorithms are developed with the addition of data acknowledge message multicasting. Application of the modified algorithms yields more performance improvements in some codes. Architectural enhancements to the network interface and directory of the SOME-Bus multiprocessor to support the contention resolution algorithms are also presented.

Distributed Shared Memory (DSM) systems have gained popular acceptance by combining the scalability and low cost of distributed system with the ease of use of single address space. Many new hardware DSM and software DSM systems have been proposed in recent years. In general, benchmarking is widely used to demonstrate the performance advantages of new systems. However, the common method used to summarize the measured results is the arithmetic mean of ratios, which is incorrect in some cases. Furthermore, many published papers list a lot of data only, and do not summarize them effectively, which confuse users greatly. In fact, many users want to get a single number as conclusion, which is not provided in old summarizing techniques. Therefore, a new data-summarizing technique based on confidence interval is proposed in this paper. The new technique includes two data-summarizing methods: (1) paired confidence interval method; (2) unpaired confidence interval method. With this new technique, it is concluded that at some confidence one system is better than others. Four examples are shown to demonstrate the advantages of this new technique. Furthermore, with the help of confidence level, it is proposed to standardize the benchmarks used for evaluating DSM systems so that a convincing result can be got. In addition, the new summarizing technique fits not only for evaluating DSM systems, but also for evaluating other systems, such as memory system and communication systems.

The single factor limiting the harnessing of the enormous computing power of clusters for parallel computing is the lack of appropriate software. Present cluster operating systems are not built to support parallel computing – they do not provide services to manage parallelism. The cluster operating environments that are used to assist the execution of parallel applications do not provide support for both Message Passing (MP) or Distributed Shared Memory (DSM) paradigms. They are only offered as separate components implemented at the user level as library and independent servers. Due to poor operating systems users must deal with computers of a cluster rather than to see this cluster as a single powerful computer. A Single System Image of the cluster is not offered to users. There is a need for an operating system for clusters. We claim and demonstrate that it is possible to develop a cluster operating system that is able to efficiently manage parallelism, support Message Passing and DSM and offer the Single System Image. In order to substantiate the claim the first version of a cluster operating system, called GENESIS, that manages parallelism and offers the Single System Image has been developed.

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.

High speed networks and rapidly improving microprocessor performance make the network of workstations an extremely important tool for parallel computing in order to speedup the execution of scientific applications. Shared memory is an attractive programming model for designing parallel and distributed applications, where the programmer can focus on algorithmic development rather than data partition and communication. Based on this important characteristic, the design of systems to provide the shared memory abstraction on physically distributed memory machines has been developed, known as Distributed Shared Memory (DSM). DSM is built using specific software to combine a number of computer hardware resources into one computing environment. Such an environment not only provides an easy way to execute parallel applications, but also combines available computational resources with the purpose of speeding up execution of these applications. DSM systems need to maintain data consistency in memory, which usually leads to communication overhead. Therefore, there exists a number of strategies that can be used to overcome this overhead issue and improve overall performance. Strategies as prefetching have been proven to show great performance in DSM systems, since they can reduce data access communication latencies from remote nodes. On the other hand, these strategies also transfer unnecessary prefetching pages to remote nodes. In this research paper, we focus on the access pattern during execution of a parallel application, and then analyze the data type and behavior of parallel applications. We propose an adaptive data classification scheme to improve prefetching strategy with the goal to improve overall performance. Adaptive data classification scheme classifies data according to the accessing sequence of pages, so that the home node uses past history access patterns of remote nodes to decide whether it needs to transfer related pages to remote nodes. From experimental results, we can observe that our proposed method can increase the accuracy of data access in effective prefetch strategy by reducing the number of page faults and misprefetching. Experimental results using our proposed classification scheme show a performance improvement of about 9---25% over the same benchmark applications running on top of an original JIAJIA DSM system.

Workload distribution is an important issue for user applications to obtain a good performance on software distributed shared memory (DSM) systems. The workload distribution methods that were previously proposed for DSM systems always considered only CPU power and data sharing but ignored memory availability of processors. However, the cost of memory accesses is also an important factor of determining program performance especially when processors have not sufficient memory space to cache all the data needed by program threads. Therefore, the past methods might make wrong decisions in workload distribution to degrade program performance. To resolve this problem, this paper is dedicated to adding memory resource consideration into workload distribution for software DSM systems. Our experimental result shows that one workload distribution method with memory resource consideration is indeed more effective for minimizing the execution time of DSM applications than another without memory resource consideration.

The demand for cloud computing has increased manifold in the recent past. More specifically, on-demand computing has seen a rapid rise as organizations rely mostly on cloud service providers for their day-to-day computing needs. The cloud service provider fulfills different user requirements using virtualization - where a single physical machine can host multiple Virtual Machines. Each virtual machine potentially represents a different user environment such as operating system, programming environment, and applications. However, these cloud services use a large amount of electrical energy and produce greenhouse gases. To reduce the electricity cost and greenhouse gases, energy efficient algorithms must be designed. One specific area where energy efficient algorithms are required is virtual machine consolidation. With virtual machine consolidation, the objective is to utilize the minimum possible number of hosts to accommodate the required virtual machines, keeping in mind the service level agreement requirements. This research work formulates the virtual machine migration as an online problem and develops optimal offline and online algorithms for the single host virtual machine migration problem under a service level agreement constraint for an over-utilized host. The online algorithm is analyzed using a competitive analysis approach. In addition, an experimental analysis of the proposed algorithm on real-world data is conducted to showcase the improved performance of the proposed algorithm against the benchmark algorithms. Our proposed online algorithm consumed 25% less energy and performed 43% fewer migrations than the benchmark algorithms.

S1: Status of Hardware and Software for Microcomputers