Storage Codes With Flexible Number of Nodes

This paper presents flexible storage codes, a class of error-correcting codes that can recover information from number nodes. As result, one make better use the available nodes in presence unpredictable node failures and reduce data access latency. Assume system encodes $k\ell $ symbols over finite field notation="LaTeX">$\mathbb {F}$ into notation="LaTeX">$n$ nodes, each size notation="LaTeX">$\ell symbols. The code is parameterized by set tuples notation="LaTeX">$\{(R_{j},\ell _{j}): 1 \le j a\}$ , satisfying _{1} < \ell _{2} {\dots } _{a} = notation="LaTeX">$R_{1}>R_{2}> }>R_{a}$ such be reconstructed any notation="LaTeX">$R_{j}$ accessing _{j}$ symbols, for notation="LaTeX">$1 a$ . In other words, allows decoding to accommodate variance time Code constructions are presented different scenarios, including LRC (locally recoverable) PMDS (partial MDS) MSR (minimum regenerating) codes. We analyze latency perform simulations on Amazon clusters show efficiency