Page 17 - profile2012.indd

Research Laboratories 研究群

電腦系統實驗室

Computer Systems Laboratory

Research Faculty
Research Faculty

Jan-Jan Wu Yuan-Hao Chang Sheng-Wei Chen PeiZong Lee Chien-Min Wang Pen-Chung Yew
Research Fellow Assistant Research Fellow Associate Research Fellow Research Fellow Associate Research Fellow Correspondence Research Fellow

Group Profile
The Computer Systems Lab was established in 2009. Its primary single-core to multicores. One of the goals is to take binary middleware. Such an environment usually faces unpredictable failures. Thus,
research areas include multicore systems, virtualization, system codes from multiple instruction-set architectures and translate fault tolerance becomes a critical requirement. QoS support is the ability to
software for cloud computing and related applications, and them to another set of instruction-set architectures on multi- guarantee a certain level of performance to an application. It is especially
storage and operating system designs for embedded systems. cores. All of such binary manipulations are under one unified important for Clouds because re-
dynamic compilation framework. Furthermore, this occurs at sources are usually virtualized and
1. System Design Support for Multicore Systems the binary-code level, instead of using high-level language shared. Finally, in a typical data Computer systems research
virtual machines such as Java and C#. It affords much higher center, storage systems consume
This research effort focuses on the architecture and system sup- transparency and portability. more than 28% of total energy con- aims to build innovative
port needed for various new system features. Designing such sumption. Therefore, saving energy
multicore systems requires extensive simulations at the sys- 3. Design and Implementation of Cloud Infrastructure for Net- of storage systems is critical for cost systems of sufficient scale,
tem level. In order to shorten critical time-to-market, software work Virtual Environments (co-funded by CITI and IIS) reduction and environment protec-
development needs to proceed tion. performance and usability
on simulators at the same time Today, to provide a quality gaming to directly impact the
hardware is being designed and experience to players, MMORPG op- 5. Storage System Designs for Em- Architecture of Storage Systems on Embedded Systems
implemented. Hence, speeding erators tend to overly provision game bedded Systems practice of computing.
up these simulations will facilitate servers. The dedicated game servers
faster design time and produce could be attributed to 40% of the Embedded systems usually adopt flash memory as their storage media. Be- We believe that the effort
better system designs. Our ap- operation cost of an MMOG. Despite cause of the advance of manufacturing technologies, flash-based storage involved in building realistic
proach is to use existing multicore of the mature virtualization technol- systems face critical challenges on performance and reliability problems.
platforms to simulate new multi- ogy, there is still no ready-to-use so- Meanwhile, existing file systems do not utilize the characteristics of flash systems is amply rewarded
core designs by exploiting their lution for a cloud-aware computation memory, and embedded systems are usually battery-backed devices that
inherent thread-level parallelism. platform for MMOGs. One of our re- might lose data due to accidental power losses. Thus, we designed new by the insights gained and
To speed up such simulations and search efforts is to design and imple- cache systems for general-purpose file systems with new non-volatile stor- the impact achieved.
obtain accurate simulation re- ment a cloud-based infrastructure for age media to prevent from data losses upon power losses and to improve
sults, various existing techniques MMOGs by utilizing virtualization and the file-system’s efficiency over flash storage devices. Now we are develop-
for single cores are extended to zone partitioning techniques. Our ing new energy-saving technologies for embedded storage systems so as to
multicores. A multi-threaded binary translation system for virtualization research will provide a fundamental realize the concept of “green storage” in embedded systems.
understanding and standard meas-
2. System Support for Virtualization urement benchmarks in this area. We expect that our platform 6. Operating System Designs for Embedded Systems
will reduce significant cost and energy consumption for operat-
Virtualization is an important technology for multicores and ing MMOGs, and also enable indie game developers to focus on As the fast-growing complexity of the operating systems on embedded
cloud computing. It allows applications running on such sys- game design. Moreover, our research results will firm the base systems requires more powerful hardware resources, the development of
tems to be agnostic about the underlying platforms. This re- of gaming on cloud computing and benefit future researches. embedded systems faces critical challenges in energy consumption and sys-
search effort focuses on the core technologies, in particular, dy- tem performance. Meanwhile, battery-backed embedded systems are usu-
namic compilation techniques for binary translation and binary 4. Distributed File Systems for Clouds ally forced to shut down for energy saving when the battery is running low.
optimization targeting multicore systems. The research team is Hence, we developed new hibernation techniques to reduce the resuming
exploring new ways of extending compilation techniques for Distributed file systems for Clouds integrate a large number of time of operating systems by stealing the idle time of storage devices, and
storage devices and provide a virtualized huge storage space. to dynamically adjust the CPU frequencies and the device’s power modes for
However, the design of such systems faces several challenges. energy saving according to the system’s runtime information. Now we are in-
In particular, we are interested in fault tolerance, QoS support, vestigating fast hibernation technologies with finer hibernation granularities
and energy conservation. Storage resources are distributed in and designing advanced energy-efficient resource scheduling algorithms to
different locations, connected by networks, and managed by further improve the efficacy of embedded systems.
研究群
研究群
Research Laboratories
16 Research Laboratories
16
17
17