As the big data revolution continues to expand rapidly, problems of treatment and protection are also expanding. Combining Advanced ECC and geographic and hardware redundancy can reduce the possibility of data loss compared to asteroid attack. The overhead is 40%, much lower than the triple redundancy used in some environments. Storage in the cloud
However, if data redundancy is compromised due to hardware disk, server, or data center failure, redundancy must be restored. As the size of the data restoration increases, the cost of repair becomes important.
The main repair costs in distributed storage systems are bandwidth and computational complexity. Bandwidth is because data must be interconnected to pass data from the source data to the repaired data. Calculate because lost data is mathematically protected and you need to reconstruct the calculation.
There is code such as Minimum Distance Separation Code (MDS) suitable for fault tolerance and overload capacity. However, these codes have high bandwidth cost.
On the other hand, the pyramid code is an extra of non-MDS code optimized to minimize the number of contacted nodes, and reduces the bandwidth requirement for data reconstruction. Researchers have developed code that minimizes bandwidth or optimizes storage efficiency.
In a recent article, Swap, Norwegian and French researchers have provided solutions for efficient storage and low cost repair problems. "Code Construction of Distributed Storage with Low Bandwidth Repair and Low Repair Complexity"
. . . It offers a non-MDS ECC family that provides low repair bandwidth and low repair complexity while maintaining relatively low field size and variable fault tolerance. In particular, we propose a systematic code structure based on two classes of parity symbols.
They are Parity knot. The first class consists of MDS code with "piggy back" added to part of the code symbol and is intended to provide ECC.
The second class of parity nodes uses block codes to which parity symbols are simply added. This class is aimed at reducing repair bandwidth and complexity by repairing faulty symbols in the node.
Testing these codes showed that the repair bandwidth is reduced by 30 to 64% compared to the MDS code. Network bandwidth is usually the most important so …
Hope you like the news Minimizing repair costs in cloud storage. Stay Tuned For More Updates 🙂