Wei Wang

Wei Wang

Auto-generated profile, based on publicly available information

Wei Wang

Hong Kong University of Science and Technology, Hong Kong, Hong Kong

Research Focus & Metrics

Academic Career Length

18Years (2009-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
9 October 2014 in IEEE Transactions on Parallel and Distributed Systems
Verified
Journal Article
Closed access
WWWei Wang
Wei Wang
BLBen Liang
Ben Liang
BLBaochun Li
Baochun Li
We study the multi-resource allocation problem in cloud computing systems where the resource pool is constructed from a large number of heterogeneous servers, representing different points in the configuration space of resources such as processing, memory, and storage. We design a multi-resource allocation mechanism, called DRFH, that generalizes the notion of Dominant Resource Fairness (DRF) from a single server to multiple heterogeneous servers. DRFH provides a number of highly desirable properties. With DRFH, no user prefers the allocation of another user; no one can improve its allocation without decreasing that of the others; and more importantly, no coalition behavior of misreporting resource demands can benefit all its members. DRFH also ensures some level of service isolation among the users. As a direct application, we design a simple heuristic that implements DRFH in real-world systems. Large-scale simulations driven by Google cluster traces show that DRFH significantly outperforms the traditional slot-based scheduler, leading to much higher resource utilization with substantially shorter job completion times.

26 January 2023 in IEEE Transactions on Parallel and Distributed Systems
Verified
Journal Article
Closed access
MLMingzhe Li
Mingzhe Li
WWWei Wang
Wei Wang
AZAndrew Zhang
Andrew Zhang
Blockchain sharding has been increasingly used to improve blockchain systems’ performance, in which a blockchain is split into multiple smaller, disjoint shards. In practice, however, sharding can only achieve limited throughput and latency improvement, especially for the user-perceived transaction confirmation delay. The performance degradation is believed to be caused by the cross-shard transactions. However, we show, through comprehensive system deployment and measurement studies, that the main culprit is the imbalanced transaction load on different blockchain shards. To address this problem, we propose a novel sharding system, called LB-Chain, which dynamically balances the transaction load on different shards by periodically migrating active accounts from heavily-loaded shards to less-loaded ones. We have implemented a prototype of LB-Chain, and evaluated its performance through large-scale blockchain deployment using real-world transaction traces. Extensive experiments confirm that LB-Chain significantly boosts sharding performance, reducing the transaction confirmation delays by up to 90% while increasing the transaction throughput by more than 10%. The delay difference between different accounts is also reduced dramatically, leading to improved fairness in the system.

1 April 2014
Verified
Conference Paper
Closed access
WWWei Wang
Wei Wang
BLBaochun Li
Baochun Li
BLBen Liang
Ben Liang
We study the multi-resource allocation problem in cloud computing systems where the resource pool is constructed from a large number of heterogeneous servers, representing different points in the configuration space of resources such as processing, memory, and storage. We design a multi-resource allocation mechanism, called DRFH, that generalizes the notion of Dominant Resource Fairness (DRF) from a single server to multiple heterogeneous servers. DRFH provides a number of highly desirable properties. With DRFH, no user prefers the allocation of another user; no one can improve its allocation without decreasing that of the others; and more importantly, no user has an incentive to lie about its resource demand. As a direct application, we design a simple heuristic that implements DRFH in real-world systems. Large-scale simulations driven by Google cluster traces show that DRFH significantly outperforms the traditional slot-based scheduler, leading to much higher resource utilization with substantially shorter job completion times.

1 July 2013
Verified
Conference Paper
Closed access
WWWei Wang
Wei Wang
DNDi Niu
Di Niu
BLBaochun Li
Baochun Li
BLBen Liang
Ben Liang
Infrastructure-as-a-Service clouds offer diverse pricing options, including on-demand and reserved instances with various discounts to attract different cloud users. A practical problem facing cloud users is how to minimize their costs by choosing among different pricing options based on their own demands. In this paper, we propose a new cloud brokerage service that reserves a large pool of instances from cloud providers and serves users with price discounts. The broker optimally exploits both pricing benefits of long-term instance reservations and multiplexing gains. We propose dynamic strategies for the broker to make instance reservations with the objective of minimizing its service cost. These strategies leverage dynamic programming and approximate algorithms to rapidly handle large volumes of demand. Our extensive simulations driven by large-scale Google cluster-usage traces have shown that significant price discounts can be realized via the broker.

1 June 2012
Verified
Conference Paper
Closed access
WWWei Wang
Wei Wang
BLBaochun Li
Baochun Li
BLBen Liang
Ben Liang
Cloud resources are usually priced in multiple markets with different service guarantees. For example, Amazon EC2 prices virtual instances under three pricing schemes -- the subscription option (a.k.a., Reserved Instances), the pay-as-you-go offer (a.k.a., On-Demand Instances), and an auction-like spot market (a.k.a., Spot Instances) -- simultaneously. There arises a new problem of capacity segmentation: how can a provider allocate resources to different categories of pricing schemes, so that the total revenue is maximized? In this paper, we consider an EC2-like pricing scheme with traditional pay-as-you-go pricing augmented by an auction market, where bidders periodically bid for resources and can use the instances for as long as they wish, until the clearing price exceeds their bids. We show that optimal periodic auctions must follow the design of m+1-price auction with seller's reservation price. Theoretical analysis also suggests the connections between periodic auctions and EC2 spot market. Furthermore, we formulate the optimal capacity segmentation strategy as a Markov decision process over some demand prediction window. To mitigate the high computational complexity of the conventional dynamic programming solution, we develop a near-optimal solution that has significantly lower complexity and is shown to asymptotically approach the optimal revenue.

27 April 2024
Verified
Conference Paper
Open access
RFRuibo Fan
Ruibo Fan
WWWei Wang
Wei Wang
XCXiaowen Chu
Xiaowen Chu
Sparse Matrix-Matrix Multiplication (SpMM) is a building-block operation in scientific computing and machine learning applications. Recent advancements in hardware, notably Tensor Cores (TCs), have created promising opportunities for accelerating SpMM. However, harnessing these hardware accelerators to speed up general SpMM necessitates considerable effort. In this paper, we undertake a comprehensive analysis of the state-of-the-art techniques for accelerating TC-based SpMM and identify crucial performance gaps. Drawing upon these insights, we propose DTC-SpMM, a novel approach with systematic optimizations tailored for accelerating general SpMM on TCs. DTC-SpMM encapsulates diverse aspects, including efficient compression formats, reordering methods, and runtime pipeline optimizations. Our extensive experiments on modern GPUs with a diverse range of benchmark matrices demonstrate remarkable performance improvements in SpMM acceleration by TCs in conjunction with our proposed optimizations. The case study also shows that DTC-SpMM speeds up end-to-end GNN training by up to 1.91× against popular GNN frameworks.