Data centers' energy requirements have increased massively in the past years. A substantial (and growing) percentage of worldwide electricity is dedicated to cloud computing, which motivates the design of energy efficient storage techniques. On the other hand, consumers of cloud storage and content centric networking demand that their content be reliably stored and readily available. Cloud storage providers today strive to meet both demands by simply replicating content throughout the storage network over multiple disks, thus increasing storage space and, consequently, energy requirements.
A large body of recent literature proposes using erasure and network coding as a more bandwidth and storage efficient way to provide reliability and accessibility of stored content. These smart storage algorithms promise to significantly cut data-center energy use, but have so far been studied and evaluated mostly in simplified theoretical settings, without taking into account all systems issues that may arise by introducing the algorithms into practical storage systems. The working group and workshop intend to bring experts in coding and queueing theory for distributed storage together with distributed storage systems experts in order to better understand and find ways to evaluate how well the practice would match the theoretical predictions of energy savings. Furthermore, other problems that arise in data storage practice and could potentially benefit from smarter algorithms will be presented.
Next: Call for Participation
Workshop Index
DIMACS Homepage