Coded caching has been widely used in the wireless network for shifting the some transmissions during the peak traffic times to the off-peak traffic times. Multi-server coded caching, which can share responsibility for the total amount of transmission in the wireless network during the peak traffic times by means of the collaboration among these servers, can be seen everywhere in our life. The three servers setting (two data servers and one parity check server) is used in practice, e.g. redundant array of independent disks-4. In this scenario, there are total N files which are equally stored in two data servers respectively and K users each of which has the memory size of M files. Each server connects to users by an independently channel. During the off-peak traffic times, two data servers place some parts of each files in each user's cache. In that time, servers do not know users' requests in future. During the peak traffic times each user just requests one file from N files. Luo et al. in 2016 proposed the first coded caching scheme for this setting. In this paper, we proposed some method that further reduces the amount of transmission in each channel when KM N is odd. This method also improves the transmission rate for systems with general multiply servers.
I. INTRODUCTIONPredominantly driven by video content demand, there is a dramatic increase in wireless traffic now. The high temporal variability of network traffic results in communication systems that are congested during peak-traffic times and underutilized during off-peak times. Caching is a natural strategy to cope with this high temporal variability by shifting some transmissions from peak to off-peak times with the help of cache memories at the network edge.Maddah-Ali and Niesen in [6] proved that coded caching does not only shift some transmissions from peak to off-peak times, but also further reduces the amount of transmission during the peak traffic times by exploiting caches to create multicast opportunities. The first caching scenario focused in [6] is: a single server containing N files with the same length connects to K users over a shared link and each user has a cache memory of size M files. During the off-peak traffic times the server places some contents to each user's cache. In this phase the server does not known what each user will require next. During the peak traffic times, each user requires a file from server randomly. Then according to each user's cache, the server sends a coded signal (XOR of some required packets) to the users such that various user demands are satisfied. The first determined coded caching scheme, which is called MN scheme in this paper, was proposed in [6]. It is worth mentioning that the broadcasted amount of MN scheme for the worst request, where all the requirements are different from each other, is at most four times larger than the lower bound when K ≤ N [2]. We denote such amount by R M N (K, M N ). So MN scheme has been extensively employed in practical scenarios, such as device to device networks [3], hie...