Locally Repairable Convertible Codes: Erasure Codes for Efficient Repair and Conversion
Francisco Maturana, K. V. Rashmi · 2023
Erasure codes are typically used in distributed storage systems in order to protect against failures and unavailabilities with low storage overhead. An important disadvantage of classic erasure codes (such as Reed-Solomon codes) is the high cost of repairing failures. Locally repairable codes (LRCs) reduce the repair cost at the cost of higher storage overhead. In practice, the parameters of LRCs are chosen based on several factors, such as failure rates, workloads, and budget constraints. However, encoded data is stored for long periods of time, and during that time these factors can vary, and thus the ideal parameters can change. The process of changing the code parameters on encoded data is called code conversion. The default approach to code conversion is to read all data, re-encode it, and write it back, which can be prohibitively expensive. To address this problem, we propose a new construction technique for designing LRCs that can perform code conversion at a lower cost than the default approach. We apply this technique to design codes for several code conversion scenarios which are of practical interest.