1: \begin{abstract}

2: In this paper, the blocklength-limited performance of a relaying system is studied, where  channels are assumed to experience  quasi-static Rayleigh fading while at the same time only the average channel state information (CSI) is available at the source. 

3: Both the physical-layer performance (blocklength-limited throughput) and the { {link}}-layer performance (effective capacity) of the relaying system are investigated.  

4: We propose a simple system operation by introducing a factor based on which we weight the average CSI and let the source determine the coding rate accordingly.

5: In particular, we prove that both the { {blocklength-limited throughput}} and the effective capacity  are quasi-concave in the weight factor. %In other words, the blocklength-limited  (either the physical-layer or the higher-layer) performance of the system can be easily optimized based on the proposed operation by choosing an appropriate weight factor.

6: %Moreover, we prove that the BL-capacity is strictly increasing in blocklength while the effective capacity is concave in the blocklength. Hence the blocklength is also an important factor in the design of QoS-support relaying systems.

7: Through numerical investigations, we show the appropriateness of our theoretical model. { {In addition, we observe that relaying is more efficient than direct transmission. Moreover,  this performance advantage of relaying under the average CSI scenario is more significant than under the perfect CSI scenario. Finally, the speed of convergence (between the blocklength-limited performance and the performance in the Shannon capacity regime) in relaying system is faster in comparison to the direct transmission under both the average CSI scenario and the perfect CSI scenario.}}

8: %Moreover, we evaluate the performance of the studied relaying system (with average CSI) in comparison to  either relaying with perfect CSI or direct transmission.

9: \end{abstract}

10: