Concurrency Control and Consistency Over Erasure Coded Data

Abstract

For over a decade, erasure codes have become an integral part of large-scale data storage solutions and data-centers. However, in commercial systems, they are, so far, used predominantly for static data. In the meanwhile, there has also been almost a decade and a half of research on mutable erasure coded data, looking at various associated issues, including update computation, concurrency control and consistency, which has led to a variety of reasonably mature techniques. In this work we aim at curating and systematizing this knowledge on managing mutable erasure coded data. We believe the time is right, both because of the richness and maturity of the literature itself, and also, given the pervasiveness of erasure codes in data-centers, because it is natural to expect a transition to accommodate mutable content using erasure coded redundancy in order to support more diverse and versatile overlying applications, while benefiting from the advantages (particularly, that of significantly lower storage overhead) of erasure codes.

Identification Codes: A Topical Review With Design Guidelines for Practical Systems

“…The choice of symbol size q = q i = 5 and dimension k i = 2 of the inner RS code determines the symbol size q o = q k i = 25 of the outer RS code. We select one of N PPM-RS2 = q k i k o = 9, 765, 625 possible IDs, that is represented in F 5 25 , namely (11,12,22,7,4) in the example illustrated in Figure 1. The ID is encoded by the outer RS code, yielding its error-correction codeword in F 24 25 .…”

“…Similar identity verification tasks arise in distributed information systems that need to maintain consistency (synchronism). For instance, distributed databases [22], [23], [24] need to maintain consistency among the different locations. Also, in multi-access edge computing (MEC), the computing instances that execute a mobile computing application need consistent application state information as the computing instances are moved among MEC servers to remain in close vicinity of the mobile application [25], [26], [27], [28].…”