R.V. Klimov - Post-graduate Student, Department «Telecommunications», Ulyanovsk State Technical University E-mail: chorrus@mail.ru

Existing models for describing distributed storage systems are only suitable for systems with Maximum-Distance-Separable codes. The paper presents an enhanced pricing model of mean time to data lost stored in a distributed storage system, in terms of arbitrary error-correcting code. In modern distributed data storage systems are widely distributed systems based of a replication. This technology is expensive to maintain a large storage array. An alternative approach is the use of error-correcting codes that reduce costs. The author shows the concept of a model a distributed storage system based on the method of concatenated coding interleaved. Additionally, this article presents analytical model for estimating the minimum traffic volume occurring during node recovery of distributed data storage system with concatenated coding. Constant growth of volumes of heterogeneous data and increasing the value of their information component leading to the increase of the role of different methods of virtualization of processing and storage. According to the latest trends in the field of information storage, every year the increasing role begin to play a variety of different services based on distributed storage systems. Such systems are built on the principle of fragmented data storage, i.e. splitting the original dataset into many separate fragments, and storing each in a separate independent from each other storage. The most important requirement for such systems is to ensure the reliability of data storage and their operational availability. In most modern systems this problem is solved by applying different methods of geo-replication. All versions of this approach are based on creating two or more copies of the source data, thereby achieving data reliability, even in the case of damage or loss of individual fragments. The obvious disadvantage of these systems is the need for storage arrays proportionally larger source. An alternative to using geo-replication is the use of different methods coding theory, in particular codes with the maximum code distance. This approach allows to significantly reduce the introduced redundancy, while ensuring the storage reliability that is comparable to geo-replication. The lack of application of coding theory is that restoring lost content nodes requires the collection of fragments located in the memory storage and the remaining components of the codeword. This leads to an increased load on the data network, which reduces data availability and increases costs of the system as a whole. The developer occurs two tasks. First, you need to assess the reliability of the system, in particular the likelihood of total data loss. Traditionally, such assessment is provided for codes with the maximum code distance, however, such solutions are not always optimal. The paper presents one approach to assessing the reliability of a system to arbitrary codes. The second objective is to optimize the traffic inside the data network, in terms of the recovery of the contents of the nodes. The author provides a solution to this problem, based on an integrated use of error correction codes with interleaving and network coding. It is also presented the method of estimating the minimum traffic restore the contents of the lost node.

 

1. Ivanichkina L.V., Neporada A.P. Model nadezhnosti raspredelennojj sistemy khranenija dannykh v uslovijakh javnykh i skrytykh diskovykh sboev // Trudy instituta sistemnogo programmirovanija RAN. 2015. T. 27. № 6. S. 253−274.
2. Patterson D.A., Gibson G., and Katz R.H. A Case for Redundant Arrays of Inexpensive Disks (RAID) // Proc. of ACM SIGMOD. 1988.
3. Mazur EH.M. Raspredelennye sistemy khranenija dannykh: analiz, klassifikacija i vybor // Perspektivy razvitija informacionnykh tekhnologijj. 2015. № 26. URL = http://cyberleninka.ru/article/n/raspredelennye-sistemy-hraneniya-dannyh-analiz-klassifikatsiya-i-vybor (13.06.2016).
4. Skljar B. Cifrovaja svjaz. M.: Radio i svjaz. 2000. 800 s.
5. Morelos-Saragosa R. Iskusstvo pomekhoustojjchivogo kodirovanija. Metody, algoritmy, primenenie. M.: Tekhnosfera. 2005. 320 s.
6. Patent RF na izobretenie № 2480918. Adaptivnyjj koder giperkoda razmernosti 3D. Gladkikh A.A., Kapustin D.A., Klimov R.V. / Bjulleten № 12. 2013.
7. Dimakis A., Godfrey P., Wu Y., Wainwright M., Ramchandran K. Network coding for distributed storage systems // IEEE Transactions on Information Theory. 2010. V. 56. № 9. P. 4539−4551.
8. Rashmi K.V., Nihar B. Shah, Vijay Kumar P. Optimal Exact-Regenerating Codes for Distributed Storage at the MSR and MBR Points via a Product-Matrix Construction // IEEE Transactions on. 2011. V. 57. № 8. P. 57(8):5227-5239.
9. Gladkikh A.A., Klimov R.V., Sorokin I.A. Metody snizhenija vnutrisetevojj nagruzki v raspredelennykh sistemakh khranenija dannykh // IKT. T. 41. № 3. 2015. S. 34−41.