Collaboration-Oriented Data Recovery for Mobile Disk Arrays Tao Xie and Abhinav Sharma Computer Science Department, San Diego State University, San Diego, USA xie@cs.sdsu.edu, sharma4@rohan.sdsu.edu Abstract Mobile disk arrays, disk arrays located in mobile data centers, are crucial for mobile applications such as disaster recovery. Due to their unusual application domains, mobile disk arrays face several new challenges including harsh operating environments, very limited power supply, and extremely small number of spare disks. Consequently, data reconstruction schemes for mobile disk arrays must be performance-driven, reliability-aware, and energy- efficient. In this paper, we develop a flash assisted data reconstruction strategy called CORE (c ollaboration-o riented re construction) on top of a hybrid disk array architecture, where hard disks and flash disks collaborate to shorten data reconstruction time, alleviate performance degradation during disk recovery. Experimental results demonstrate that CORE noticeably improves the performance and energy- efficiency over existing schemes. 1. Introduction Mobile data centers [10][16] are an alternative to conventional stationary data centers that are enclosed in buildings. They could be built on self-contained trucks, airplanes, or ships that have onboard generators, UPS, multiple high-capacity servers, and satellite Internet links [16]. For example, an NAAT MED (Mobile Emergency Datacenter) can accommodate up to 100 fully charged laptops, multiple high-performance servers, and a large capacity storage system with multiple Terabytes of data in a 20-25 ft truck [16]. Typical applications for mobile data centers include disaster recovery [6], live video broadcast [15], and homeland security [25], where high mobility and a fast large-volume data processing capability are intrinsically demanded. Apparently, mobile disk arrays are essential in these emergency-oriented applications because they can provide not only huge storage capacities but also high bandwidth. At present, a mobile disk array generally consists of an array of independent small form factor hard disks connected to a host by a storage interface like SAS (Serial-attached SCSI) [10][16]. Mobile disk arrays face several new challenges including harsh operating environments, very limited power supply, and extremely small number of spare disks, which were not experienced by their stationary counterparts before. First of all, mobile disk arrays are more prone to failures than static ones due to their severe application environments. Compared with their static counterparts mobile disk arrays generally operate in a much worse environment, which could result in a higher annual disk replacement rate [21]. Consequently, disk failures in mobile data centers become non-rare events. Next, stationary disk arrays are located in data center buildings where electrical power is guaranteed. Mobile disk arrays, however, only have very limited power supply provided by either gasoline generators or batteries. Therefore, energy-saving becomes more critical for mobile disk arrays because their energy consumption can significantly affect the life time of the entire mobile systems. Finally, very small number of spare disks can be carried in a mobile data center due to its limited space. Thus, mobile disk arrays should be able to gracefully degrade performance after disk failures occur. The new challenges demand a performance- driven, reliability-aware, and energy-efficient data reconstruction algorithm, which is executed in the presence of disk failures. Traditional data reconstruction algorithms [1][6][8][17][24][27] strived to minimize the reconstruction time and alleviate performance degradation during the recovery process using various approaches including exploiting parallelism [6][8], integrating workload access locality into reconstruction process [24], and employing excess disk capacity [27]. Although they worked well for stationary disk arrays, they are not suitable for mobile data centers for the following two reasons. First, they 2009 29th IEEE International Conference on Distributed Computing Systems 1063-6927/09 $25.00 © 2009 IEEE DOI 10.1109/ICDCS.2009.13 631