Biography
Thank you for viewing my website. I'm a post-doctoral researcher in Prof. Yong Chen's DISCL lab at Texas Tech University for the first year. I graduated from Institute of Computing Technology (ICT), Chinese Academy of Sciences (CAS) as a doctor under Porf. Dan Meng's supervisor in 2014. I was previously working as an assistant professor at Institute of Information Engineering (IIE), Chinese Academy of Sciences (CAS). I received my M.S. degree from Huazhong University of Science and Technology (HUST) in 2008 and B.E. degree from Hubei University (HBU) in 2002 in China respectively.
My current research interests include big data, distributed file system, high-performance computing, parallel and distributed computing, computer architectures and systems software design for high-performance scientific computing. During my PhD’s studies, I mainly participated in several programs including research on distributed file system for high performance computing ( Hyper Virtual File System called HVFS for "Dawning 6000" supercomputer), mass data management for distributed cloud computing platform (Clover File System called CFS functions as HDFS for Hadoop architecture) and one project (Tencent Data Warehouse called TDW) cooperation with the internet company Tencent. My now works mainly focus on multiple metadata service for big data management, distributed storage system for QEMU/KVM and data-intensive computing.
Publictions
2015
2014 and ealier
Thank you for viewing my website. I'm a post-doctoral researcher in Prof. Yong Chen's DISCL lab at Texas Tech University for the first year. I graduated from Institute of Computing Technology (ICT), Chinese Academy of Sciences (CAS) as a doctor under Porf. Dan Meng's supervisor in 2014. I was previously working as an assistant professor at Institute of Information Engineering (IIE), Chinese Academy of Sciences (CAS). I received my M.S. degree from Huazhong University of Science and Technology (HUST) in 2008 and B.E. degree from Hubei University (HBU) in 2002 in China respectively.
My current research interests include big data, distributed file system, high-performance computing, parallel and distributed computing, computer architectures and systems software design for high-performance scientific computing. During my PhD’s studies, I mainly participated in several programs including research on distributed file system for high performance computing ( Hyper Virtual File System called HVFS for "Dawning 6000" supercomputer), mass data management for distributed cloud computing platform (Clover File System called CFS functions as HDFS for Hadoop architecture) and one project (Tencent Data Warehouse called TDW) cooperation with the internet company Tencent. My now works mainly focus on multiple metadata service for big data management, distributed storage system for QEMU/KVM and data-intensive computing.
Publictions
2015
- J. Zhou, Y. Chen, W. Wang and D. Meng. MAMS: A Highly Reliable Policy for Metadata Service. Accepted to appear in the Proc. of the 44th International Conference on Parallel Processing (ICPP'15), 2015. (acceptance rate: 99/305=32.5%)
- J. Zhou, Y. Chen, X. Gu, W. Wang, and D. Meng. A Virtual Shared Metadata Storage for HDFS. Accepted to appear in the Proc. of The 10th IEEE International Conference on Networking, Architecture, and Storage (NAS'15), 2015. (acceptance rate: 28/94 = 29.8%)
2014 and ealier
- Jiang Zhou, Weiping Wang, Dan Meng, Can Ma, Xiaoyan Gu, Jie Jiang, "Research on Key Technology in Distributed File System Towards Big Data Analysis", Journal of Computer Research and Development, 51(2): 382-394, 2014. (In Chinese)
- Jiang Zhou, Jing Guo, Weiping Wang, Cuilan Du, Xiaoyan Gu, Dan Meng, "OAMS: A Highly Reliable Metadata Service for Big Data Storage", International Workshop on Big Data and Cloud Computing in Public and Business Services (DCPIS), 2013.
- Jiang Zhou, Can Ma, Jin Xiong, Weiping Wang, Dan Meng, "Highly-reliable Message-passing Mechanism for Cluster File System", International Journal of Parallel, Emergent and Distributed Systems (JPEDS), 28(6): 556-575, 2013.
- Xiaoyan Gu, Weiping Wang, Dan Meng, Xiufeng Yang, Jiang Zhou, "CI-DCG: An Efficient Data Cube Model for Supporting OLAP on Multidimensional Network", High Technology Letters, 10: 42-49, 2013. (In Chinese)
- Youwei Wang, Jiang Zhou, Can Ma, Weiping Wang, Dan Meng, Jason Kei, "Clover: A Distributed File System of Expandable Metadata Service Derived from HDFS", In the Proc. of the IEEE International Conference on Cluster Computing 2012 (Cluster), pp. 126-134, 2012.
- Jiang Zhou, Can Ma, Jin Xiong, Dan Meng, "HR-NET: A Highly Reliable Message-passing Mechanism for Cluster File System", in Proceeding of IEEE Sixth International Conference on Networking, Architecture, and Storage (NAS), pp. 364-371, 2011.
- Jiang Zhou, Jin Xiong, Can Ma, "Message-passing for High Available File System", Journal of Huazhong University of Science and Technology (Nature Science Edition), 39(SI): 139-143, 2011. (In Chinese)
Software Copyright
- Jing Xiong, Can Ma, Jiang Zhou, Zhuo Chen, Jiuyue Ma, hyper virtual file system, register number: 2011SR055421. (In Chinese)
Honors and Awards
- 2014 Second and Third Prize of Mobile Internet Application Youth Contest in CAS (TOP 15%)
- 2013 Director of Special Scholarship in ICT, CAS (TOP 3%); "Baidu" HACKATHON Competition, 1st Prize; "Baidu" Recommendation Competition, 1st Prize;
- 2012 Merit Student Awards in GUCAS (TOP 13%);
Academic Service
- External Reviewer Transactions on Computers (TC) 2014
- _External Reviewer Cloud15 2015
- External Reviewer ISPDC 2015
System and Project
- Research on Distributed File System Metadata Service for Mass Data Analysis
I mailnly design and implement a novel highly scalable and available distributed file system CFS (Clover File System) for mass data analysis. The file system adopts the approach of metadata and data separated storage with multiple servers to manage a global namespace and provides high overall I/O throughput for large files. It is based on directory sharding and consistent hashing to manage a global namespace. For distributed metatadata storage, a virtual shared storage pool (SSP) is designed for metadata synchronization and replication. Based on the SSP, a novel backp policy is proposed to improve metadata service reliability. CFS is fully compatible with HDFS interfaces, satisfies for file access pattern in big data storage and supports standard applications in Hadoop frame. This work has been published on Cluster 2013 and Journal of CRAD 2014.
I mailnly design and implement a novel highly scalable and available distributed file system CFS (Clover File System) for mass data analysis. The file system adopts the approach of metadata and data separated storage with multiple servers to manage a global namespace and provides high overall I/O throughput for large files. It is based on directory sharding and consistent hashing to manage a global namespace. For distributed metatadata storage, a virtual shared storage pool (SSP) is designed for metadata synchronization and replication. Based on the SSP, a novel backp policy is proposed to improve metadata service reliability. CFS is fully compatible with HDFS interfaces, satisfies for file access pattern in big data storage and supports standard applications in Hadoop frame. This work has been published on Cluster 2013 and Journal of CRAD 2014.
- Distributed File System for Dawning 6000 Supercomputer
I participate in the design of distributed file system HVFS
(Hyper Virtual File System), which is a cluster file system for the "Dawning 6000"
supercomputer at the data center which provides petabyte-scale storage. HVFS is
mainly designed for massive small file operations and handles large files based
on other file systems (e.g. Lustre, Panasas and PVFS2). HVFS manages all the
files by a set of metadata servers and provides a global namespace for users. The
goal is to provide high aggregate bandwidth (200GB/s) and low latency of
concurrent metadata access (10000 small files/s). My research focus on the message passing mechanism for cluster filesystem, metadata storage and client operations. Design
and implement a highly reliable message-passing mechanism (HR-NET), which
tolerates both software and hardware network failures. HR-NET detects and
recovers network failures by the fault-tolerant mechanism and priority-based
message scheduling strategy. It ensures the availability of each pair
transmission and sub-operations in a file system. The work has been published on NAS 2011 and JPEDS 2013.