Mini-Workshop on Hot Topics in Cloud Computing Systems

Friday, 29th August 2014 (PDCC Meeting Room)
Title: Mini-Workshop on Hot Topics in Cloud Computing Systems
Organiser: Xtra Computing Group @SCE, NTU
 
Introduction:

The workshop includes two invited talks from two hot system research groups in China, and two SC’14 rehearsal talks from our student/staff. The talks cover various hot system topics on cloud computing: HPC, security, virtualization, and resource allocations etc


Program:

Program:
Schedule Title Speaker
9:30-10:15 CYPRESS: Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression Dr. Jidong Zhai,
Tsinghua University, China
10:15-11:00 Concurrent and Consistent Virtual Machine Introspection with Hardware Transactional Memory Dr. Yubin Xia,
Shanghai Jiao Tong University, China
11:00-11:30 Finding Constant From Change: Revisiting Network Performance Aware Optimizations on IaaS Clouds Mr. Yifan Gong,
Nanyang Technological University
11:30-12:00 Reciprocal Resource Fairness: Towards Cooperative Multiple-Resource Fair Sharing in IaaS Clouds Dr. Haikun Liu,
Nanyang Technological University

Details:
--------------------------------------Talk No. 1-----------------------------------
Title

CYPRESS: Combining Static and Dynamic Analysis for Top-Down Communication Trace Compression

Abstract

Communication traces are increasingly important, both for parallel applications’ performance analysis/optimization, and for designing next-generation HPC systems. Meanwhile, the problem size and the execution scale on supercomputers keep growing, producing prohibitive volume of communication traces. To reduce the size of communication traces, existing dynamic compression methods introduce large compression overhead with the job scale. We propose a hybrid static-dynamic method that leverages information acquired from static analysis to facilitate more effective and efficient dynamic trace compression. Our proposed scheme, CYPRESS, extracts a program communication structure tree at compile time using inter-procedural analysis. This tree naturally contains crucial iterative computing features such as the loop structure, allowing subsequent runtime compression to “fill in”, in a “top-down” manner, event details into the known communication template. Results show that CYPRESS reduces intra-process and inter-process compression overhead up to 5x and 9x respectively over state-of-the-art dynamic methods, while only introducing very low compiling overhead.

Short Biography

Jidong Zhai received the bachelor's degree in computer science from University of Electronic Science and Technology of China in 2003, and the Ph.D. degree in computer science from Tsinghua University in 2010. He is currently an assistant professor in the Department of Computer Science and Technology, Tsinghua University. His research focuses on high performance computing, compiler optimization, performance analysis and optimization of large-scale parallel applications. He is a recipient of Siebel Scholar and CCF outstanding doctoral dissertation award.

--------------------------------------Talk No. 2-----------------------------------
Title

Concurrent and Consistent Virtual Machine Introspection with Hardware Transactional Memory

Abstract

Virtual machine introspection, which provides tamper-resistant, high-fidelity “out of the box” monitoring of virtual machines, has many prominent security applications including VM-based intrusion detection, malware analysis and memory forensic analysis. However, prior approaches are either intrusive in stopping the world to avoid race conditions between introspection tools and the guest VM, or providing no guarantee of getting a consistent state of the guest VM. Further, there is currently no effective means for timely examining the VM states in question.

In this paper, we propose a novel approach, called TxIntro, which retrofits hardware transactional memory (HTM) for concurrent, timely and consistent introspection of guest VMs. Specifically, TxIntro leverages the strong atomicity of HTM to actively monitor updates to critical kernel data structures. Then TxIntro can mount introspection to timely detect malicious tampering. To avoid fetching inconsistent kernel states for introspection, TxIntro uses HTM to add related synchronization states into the read set of the monitoring core and thus can easily detect potential inflight concurrent kernel updates. We have implemented and evaluated TxIntro based on Xen VMM on a commodity Intel Haswell machine that provides restricted transactional memory (RTM) support. To demonstrate the effectiveness of TxIntro, we implemented a set of kernel rootkit detectors using TxIntro. Evaluation results show that TxIntro is effective in detecting these rootkits, and is efficient in adding negligible performance overhead.

Short Biography

Yubin Xia received the diploma degree in software school, Fudan University, Shanghai, China, in 2004, and the Ph.D. degree in computer science and technology from Peking University, Beijing, China, in 2010. After being a post-doc in Fudan University during 2010 to 2012, he's now an assistant professor in Institute of Parallel and Distributed Systems, Shanghai Jiao Tong University, Shanghai, China since September 2012. His research interests include the area of computer architecture, system software, and security.

--------------------------------------Talk No. 3-----------------------------------
Title

Finding Constant From Change: Revisiting Network Performance Aware Optimizations on IaaS Clouds

Abstract

Network performance aware optimizations have long been an effective approach to optimize distributed applications on traditional network environments. But the assumptions are no longer valid on IaaS clouds. The virtualization technology hides network topology from users, and direct use of network performance measurements may not represent the long-term performance. We propose to decouple the constant component from the dynamic network performance while minimizing the difference between network performance and constant component by a mathematical method called RPCA. We use the constant component to guide the network performance aware optimizations. We demonstrate the efficiency and effectiveness of our approach by adopting network aware optimizations for collective communications of MPI and generic topology mapping. Moreover, we implement two real-world applications, N-body and conjugate gradient (CG), and compare the performance efficiency with our approach. Our experiments on Amazon EC2 and simulations demonstrate significant performance improvement on guiding the network performance aware optimizations.

--------------------------------------Talk No. 4-----------------------------------
Title

Reciprocal Resource Fairness: Towards Cooperative Multiple-Resource Fair Sharing in IaaS Clouds

Abstract

This paper presents VMbuddy, a cooperative resource management system for infrastructure-as-a-service (IaaS) cloud. VMbuddy advocates a group of cloud tenants to buy resource capacity in bulk and share the resource pool in the form of virtual machines(VMs). However, fine-grained resource sharing, especially for multiple types, poses several challenging problems, such as sharing incentive, fairness, and free-riding. To address these problems, we propose Reciprocal Resource Fairness (RRF), a generalization of max-min fairness to multiple types of resource. RRF is implemented in two complementary and hierarchical mechanisms: inter-tenant resource trading and intra-tenant weight adjustment. We implement VMbuddy in Xen platform. The experimental results show VMbuddy is promising for both cloud providers and tenants. For cloud providers, VMbuddy improves VM density and their revenue by 2.2X compared to the current IaaS cloud models. For tenants, VMbuddy delivers better application performance and 95% economic fairness among multiple tenants.