Parallel and distributed file systems are widely used to provide high throughput in high-performance computing and Cloud computingsystems. To increase the parallelism, I/O requests are partitioned into multiple sub-r...
详细信息
Parallel and distributed file systems are widely used to provide high throughput in high-performance computing and Cloud computingsystems. To increase the parallelism, I/O requests are partitioned into multiple sub-requests (or `flows') and distributed across different data nodes. The performance of file systems is extremely poor if data nodes have highly unbalanced response time. Client-side caching offers a promising direction for addressing this issue. However, current work has primarily used client-side memory as a read cache and employed a write-through policy which requires synchronous update for every write and significantly under-utilizes the client-side cache when the applications are write-intensive. Realizing that the cost of an I/O request depends on the struggler sub-requests, we propose a cost-aware client-side file caching (CCFC) strategy, that is designed to cache the sub-requests with high I/O cost on the client end. This caching policy enables a new trade-off across write performance, consistency guarantee and cache size dimensions. Using benchmark workloads MADbench2, we evaluate our new cache policy alongside conventional write-through. We find that the proposed CCFC strategy can achieve up to 110% throughput improvement compared to the conventional write-through policies with the same cache size on an 85-node cluster.
Mitogen-activated protein kinase(MAPK) cascade,consisting of three Ser/Thr protein kinases,namely,MAPKKK,MAPKK and MAPK,is a universal module in signal transduction pathways in *** proteins have been implicated in div...
详细信息
Mitogen-activated protein kinase(MAPK) cascade,consisting of three Ser/Thr protein kinases,namely,MAPKKK,MAPKK and MAPK,is a universal module in signal transduction pathways in *** proteins have been implicated in diverse cellular processes including cell growth,proliferation,differentiation,survival, development and in responses to a diversity of environmental stimuli including cold, heat,reactive oxygen species,UV,drought and pathogen attack.
Recent advance of virtualization technology provides a new approach to check-point/restart at the virtual machine(VM) *** contrast to traditional process-level checkpointing,checkpointing at the virtualization layer b...
详细信息
Recent advance of virtualization technology provides a new approach to check-point/restart at the virtual machine(VM) *** contrast to traditional process-level checkpointing,checkpointing at the virtualization layer brings up several advantages,such as compatibility,transparence,flexibility and ***,because the virtualization layer has little semantic knowledge about the operation system and the applications running atop,VM-layer checkpointing requires saving the entire operating system state rather than a single *** overhead may render the approach *** reduce the size of VM checkpoint,in this paper we propose a page eviction scheme and an incremental checkpointing mechanism to avoid saving unnecessary VM pages in the *** keep the system online transparently,we propose a live checkpointing mechanism by saving the memory image in a copy-on-write(COW) *** implement the performance optimization mechanisms in a prototype system,called *** results with a group of representative applications show that our page eviction scheme and incremental checkpointing can significantly reduce the checkpoint file size by up to 87% and shorten the total checkpointing/restart time by a factor of up to 71%,in comparison with the Xens default checkpointing *** observed application downtimes due to checkpointing can be reduced to as small as 300 ms.
Recently, graphics processing units (GPUs) have opened up new opportunities for speeding up general-purpose parallel applications due to their massive computational power and up to hundreds of thousands of threads ena...
详细信息
ISBN:
(纸本)9781450319003
Recently, graphics processing units (GPUs) have opened up new opportunities for speeding up general-purpose parallel applications due to their massive computational power and up to hundreds of thousands of threads enabled by programming models such as CUDA. However, due to the serial nature of existing micro-architecture simulators, these massively parallel architectures and workloads need to be simulated sequentially. As a result, simulating GPGPU architectures with typical benchmarks and input data sets is extremely time-consuming. This paper addresses the GPGPU architecture simulation challenge by generating miniature, yet representative GPGPU kernels. We first summarize the static characteristics of an existing GPGPU kernel in a profile, and analyze its dynamic behavior using the novel concept of the divergence flow statistics graph (DFSG). We subsequently use a GPGPU kernel synthesizing framework to generate a miniature proxy of the original kernel, which can reduce simulation time significantly. The key idea is to reduce the number of simulated instructions by decreasing per-thread iteration counts of loops. Our experimental results show that our approach can accelerate GPGPU architecture simulation by a factor of 88X on average and up to 589X with an average IPC relative error of 5.6%.
Map/Reduce style data-parallel computation is characterized by the extensive use of user-defined functions for data processing and relies on data-shuffling stages to prepare data partitions for parallel computation. I...
The simulation grid has become a very important research topic. In order to ensure the QoS of resource management, we present a control mechanism in simulation grid, which is based on the prediction methods. In the co...
详细信息
Transactional memory (TM) is a parallel programming concept which reduces challenges in parallel programming. Existing distributed transactional memory system consumes too much bandwidth and brings high latency. In th...
详细信息
Due to great advances in computing and Internet technologies, organizations have been enabled to collect and generate a large amount of data. Most of these organizations tend to analyze their data to discover new patt...
详细信息
Due to great advances in computing and Internet technologies, organizations have been enabled to collect and generate a large amount of data. Most of these organizations tend to analyze their data to discover new patterns. Usually, analyzing such amount of data requires huge computational power and storage facilities that may not be available to these organizations. Cloud computing offers the best way to solve this problem. Storing the private data of different organizations in the same cloud server enhances the mining process, but at the same time, raises privacy concerns. Therefore, it is highly recommended to support privacy preserving data mining algorithms in the cloud environment. This paper introduces an efficient and accurate cryptography-based scheme for mining the cloud data in a secure way without loss of accuracy. Specifically, we address the problem of K-nearest neighbor (KNN) classification over horizontally distributed databases without revealing any unnecessary information. We have utilized the recently developed cryptography primitive, order preserving symmetric encryption (OPSE), to integrate securely the local classifications at a lower cost than the previously presented privacy preserving data mining schemes. Empirical results on real datasets demonstrate that the proposed scheme has similar performance with the naive mining systems in terms of classification accuracy.
Advances in cloud computing and Internet technologies have pushed more and more data owners to outsource their data to remote cloud servers to enjoy with huge data management services in an efficient cost. However, de...
详细信息
Advances in cloud computing and Internet technologies have pushed more and more data owners to outsource their data to remote cloud servers to enjoy with huge data management services in an efficient cost. However, despite its technical advances, cloud computing introduces many new security challenges that need to be addressed well. This is because, data owners, under such new setting, loss the control over their sensitive data. To keep the confidentiality of their sensitive data, data owners usually outsource the encrypted format of their data to the untrusted cloud servers. Several approaches have been provided to enable searching the encrypted data. However, the majority of these approaches are limited to handle either a single keyword search or a Boolean search but not a multikeyword ranked search, a more efficient model to retrieve the top documents corresponding to the provided keywords. In this paper, we propose a secure multi-keyword ranked search scheme over the encrypted cloud data. Such scheme allows an authorized user to retrieve the most relevant documents in a descending order, while preserving the privacy of his search request and the contents of documents he retrieved. To do so, data owner builds his searchable index, and associates with each term document with a relevance score, which facilitates document ranking. The proposed scheme uses two distinct cloud servers, one for storing the secure index, while the other is used to store the encrypted document collection. Such new setting prevents leaking the search result, i.e. the document identifiers, to the adversary cloud servers. We have conducted several empirical analyses on a real dataset to demonstrate the performance of our proposed scheme.
Virtualization can divide or aggregate the underlying resource flexibly, and it attracts attention from both academic and industry in recent years. Guest operating systems in virtual machines can be various, and the s...
详细信息
Virtualization can divide or aggregate the underlying resource flexibly, and it attracts attention from both academic and industry in recent years. Guest operating systems in virtual machines can be various, and the states of virtual machines changes all the time. Due to the complexity of virtual computing environment, it brings tremendous challenges for security monitoring. Existing monitoring mechanism can not adapt to the dynamic and diversity of virtual machines. Therefore, a comprehensive monitoring framework, named ComMon, is proposed in this paper, which implements comprehensive monitoring from three aspects: network, process, and file. It has the characteristics of real-time, transparency and generality.
暂无评论