Making distributed storage systems reliable is an important challenge. Simple replication may cause severe storage overhead when individual components of The system are very unreliable. Using erasure codes is a promising solution for This problem, but it presents its own challenges; it makes The design of such a system very complex, and it presents The problem of reparation. Network coding has been suggested To be used in The communication in These networks To help reduce overhead. However, using random network coding as -not besides -erasure coding would be an even more promising field To investigate; such a system would have a simple design, need little or no centralization, and reparation of The system could be much simpler Than it is in other erasure coding schemes. The first step on This path is To investigate whether network coding can achieve such a performance That it is a feasible alternative To other erasure codes. This paper presents our experiences about The realization of random network coding based on The discrete logarithm of The finite field. We discuss possible performance optimizations for such a system, and provide performance measurement results focusing on data storage scenarios.