Research status:

Research fields in our research group include but are not limited to:

Non-volatile memory (NVM) systems

In the field of NVM systems and technologies, a major problem is that legacy storage architectures and softwares are mostly tuned for conventional storage media (e.g. hard disk drive). Thus when NVMs replace traditional storage in modern computer systems, they tend to underperform mainly due to discordance between traditional softwares, systems and new hardwares. Another concern is the great room for improvement on the performance and endurance of NVMs. We have been conducting research on storage architecture, file system and related technologies on NVM to address these issues. The continuing efforts and research results of our team are recognized by publications on international conferences (including FAST’2014/2013, USENIX ATC’2016, EuroSys’2016, MICRO-46, ISCA’2015, DAC’2016/2015, DSN’2014, MSST’2016/2015, IPDPS’2014, ICCD’2014/2013, DATA’2014, ICPP’2016, CF’2015, GLSVLSI’16, ASP-DAC’2016, NVMSA’2016/2015/2014) and journals (including IEEE TOC 2016/2015, ACM TOS 2016). Our paper "Blurred Persistence in Transactional Persistent Memory" was a nominee for Best Paper Award in MSST 2015.

Network/cloud/big data storage systems
Parallel/distributed processing

In the field of network/cloud/big data storage systems, our research focus is around the performance, scalability, compatibility, usability, security and manageability of large scale storage systems. We have been undertaking research on storage virtualization, online scaling, metadata management, caching strategy, hierarchical storage, SAN-based parallel file system, cloud storage security, etc.. Our research projects also yielded multiple industrial-grade systems. Among which the TH-MSNS (Tsinghua University Mass Storage Network System, independent intellectual property rights) has over 100 installations adopted by departments and industries including audit, public security, telecommunication, petroleum, higer education, etc. by the year 2006. Other sub-systems including storage management (TH-MSIS), distributed virtual storage (TH-DVSM), out-of-band virtual storage (TH-VSM), backup (TH-Backup), snapshot (TH-Snapshot), mirroring (TH-Mirror), SAN-based parallel file storage (Redbud), hierachical storage (AIP), cloud storage security (Corslet) have also seen application telecommunication, education, etc.. The above mentioned research projects were awarded 2nd class National Science and Technology Progress Prize in 2007 and 2nd class State Technological Invention Prize in 2015, etc.. Related research results also appeared in international conferences including IPDPS’2016, IWQos’2014/2013, ICPP’2012, CLUSTER’2012, ASP-DAC’2013, SISW’2008, SPEED’2008, MSST’2006, ICCS‘2006/2005, NPC’2005/2004, CCGrid’2004, as well as international journals, including IEEE TOC 2016/2014/2010/2007/2005, ACM TOS 2010/2007, IEEE TPDS 2016/2015/2014, and JPDC 2014.

Data storage reliability

In the field of storage reliability, we have conducted research on the fault tolerance (1-, 2- and multi-fault tolerant) of storage erasure codes, coding efficiency and storage utilization. We proposed and implemented several efficient storage erasure codes to boost the reliability of storage systems. Our publications in this research field include conference papers like DSN’2016/2014, SRDS’2016/2015/2014, IPDPS’2015, ICPP’2015/2013/2012, IWQos’2016, ISIT’2010 and journal articles like IEEE TOC 2016/2010, IEEE TPDS 2016, and ACM TOS 2012/2009. Our paper "Seek-Efficient I/O Optimization in Single Failure Recovery for XOR-Coded Storage Systems" was one of the Nominated Best Papers in SRDS 2015.