To investigate the advantages of mixed carbon source over a single one in deep denitrification, sodium acetate, glucose and their mixture were used as carbon sources in present study. Denitrification performance, effluent pH, microbial community and carbon source cost were taken into account. With the same influent NO3–-N concentration of 50 mg/L and the same C/N ratio of 1.5, the NO3–-N removal rate with the mixed carbon source (96.53%) was slightly lower than that with sodium acetate (98.15%), but significantly higher than that with glucose (74.69%). The specific denitrification rates of the sodium acetate, glucose and sodium acetate/glucose reactor were 47.7, 29.7 and 45.4 mg N/g VSS d, respectively. The effluent pH with sodium acetate varied in the range of 9.13–9.60, exceeding the discharge standard limit of 9.0, whereas the sodium acetate/glucose reactor could keep pH in the range of 7.80–8.23. The 16S rRNA gene-based high-throughput sequencing revealed that carbon sources determined the microbial community structure and the sludge Shannon index with the mixed carbon source was the highest. Furthermore, cost estimation indicated that the mixed carbon source was the cheapest. This study is significant as it tests reasonable selection of carbon sources for deep denitrification in practice.