In this letter, we consider an intelligent reflecting surface (IRS)-aided wireless relaying system, where a decode-and-forward relay (R) is employed to forward data from a source (S) to a destination (D), aided by M passive reflecting elements. We consider two practical IRS deployment strategies, namely, single-IRS deployment where all reflecting elements are mounted on one single IRS that is deployed near S, R, or D, and multi-IRS deployment where the reflecting elements are allocated over three separate IRSs which are deployed near S, R, and D, respectively. Under the line-of-sight (LoS) channel model, we characterize the capacity scaling orders with respect to an increasing M for the IRS-aided relay system with different IRS deployment strategies. For single-IRS deployment, we show that deploying the IRS near R achieves the highest capacity as compared to that near S or D. While for multi-IRS deployment, we propose a practical cooperative IRS passive beamforming design which is analytically shown to achieve a larger capacity scaling order than the optimal single-IRS deployment (i.e., near R) when M is sufficiently large. Numerical examples are provided, which validate our theoretical results.