Distributed Mobile Storage Systems:

The PersonalRAID-0, Skunk, and Segank Systems




This non-uniformly connected world is the reality today and it is continuing to evolve. There are three aspects of this development. First, low-cost short-range wireless technologies, such as 802.11 and Bluetooth, are proliferating, which allow mobile elements in a small neighborhood to be spontaneously connected with each other at a level of quality that is quite good. Second, when a fast WiFi ``gateway'' into the Internet is not available, pervasive but low-quality wireless connectivity in the wide area using technologies such as cellular modems has become very affordable. Third, stationary storage elements are becoming increasingly wired and they are ``always on'' the network. These may include not only computers in offices and server rooms, but also broadband-connected computers at home and hotels, and an increasing array of entertainment appliances such as Tivo-like personal video recorders. The connectivity quality among these devices also exhibits a high degree of variance. A typical DSL-connected home computer, for example, may only have an up-link capacity around 100 Kbps.

Despite the high variance in connectivity quality, total disconnection is (or can be) increasingly rare, as those who own BlackBerry email devices and those who experiment with Internet access on transcontinental flights are beginning to realize.

We consider some example usage scenarios. A user owns several computers. Perhaps some of them are in his office, some at his DSL-linked home, and some in an ``off-site'' office in a different city, which he occasionally visits. Some of them are desktop machines, and others are laptops and PDAs that may accompany the user when he travels.

When the user arrives at his office, some of his latest work may have been done on a laptop that he carried home the night before and is still with him. At this time, the user should not be forced to wait for all the new data to propagate from the laptop to the office desktop before he is allowed to resume work on the desktop. He should be able to see and operate on the complete and latest view of his data from his desktop immediately. Also, the user should not have to remember where the latest copy of a particular piece of data is.

The next day, the user may run into a colleague on a speeding train and the two spontaneously decide to share some files. In this case, the system should try its best to satisfy the requests using the ad hoc 802.11 link between the two laptops, and resort to a cellular modem to reach data that is only available at the office or home. On still a third day, while on the train again, the user may receive a call from a colleague in his office, requesting some files. Here, the system would instead attempt to satisfy the colleague's requests using a copy stored on the office LAN, on a DSL-linked home machine, or on the cellular modem-connected laptop on the train, in that order of preference.

To see how such a system can be built, please read the following documents.


Publications:
 
o      Sumeet Sobti, Nitin Garg, Fengzhou Zheng, Junwen Lai, Yilei Shao, Chi Zhang, Elisha Ziskind, Arvind Krishnamurthy, and Randolph Y. Wang. Segank: A Distributed Mobile Storage System. Proc. Third Conference on File and Storage Technologies. March 2004.
 
o      Fengzhou Zheng, Nitin Garg, Sumeet Sobti, Chi Zhang, Russell E. Joseph, Arvind Krishnamurthy, and Randolph Y. Wang. Considering the Energy Consumption of Mobile Storage Alternatives. IEEE Symposium on Modeling, Analysis and Simulation of Computer Systems. October 2003.
 
o      John Zedlewski, Sumeet Sobti, Nitin Garg, Fengzhou Zheng, Arvind Krishnamurthy, and Randolph Wang. Modeling Hard-Disk Power Consumption. Proc. Second Conference on File and Storage Technologies. March 2003.
 
o      Randolph Wang, Nitin Garg, Yilei Shao, Elisha Ziskind, Sumeet Sobti, Fengzhou Zheng, Junwen Lai, and Arvind Krishnamurthy. A Peer-to-Peer Mobile Storage System. Business Briefing: Data Management and Storage Technology 2002. ISBN:1-903150-73-6, World Markets Research Centre, London, October 2002.
 
o      Sumeet Sobti, Nitin Garg, Chi Zhang, Xiang Yu, Arvind Krishnamurthy, and Randolph Y. Wang. PersonalRAID: Mobile Storage for Distributed and Disconnected Computers. Proc. First Conference on File and Storage Technologies. January 2002.
 
Press:
 
o   MIT Technology Review, 02/2004 (link).
o   Technology and Telecoms Briefing, 10/2002 (link).
o   eWeek Enterprise News and Reviews, 07/2002 (link).

Support:
 
o   NSF ITR Grant: Ubiquitous Mobile Storage (CCR-0313089), 2002.
o   NSF Career Award: Low Latency I/O and Ubiquitous Storage (CCR-9984790), 1999.



© 2004    Randy Wang