Interference between independent photon sources is the key technique to realize complex quantum systems for more sophisticated applications such as multi-photon entanglement generation and quantum teleportation. Here, we report Hong-Ou-Mandel interference (HOMI) between two independent 1.55 m all-fiber photon pair sources over two 100 GHz dense wave division multiplexing (DWDM) channels, whose visibility reaches 53.2%8.4% (82.9%5.3%) without (with) back ground counts subtracted. In addition, we theoretically describe in detail the single photon spectral purity of the photon source generated in dispersion shifted fiber (DSF), simulate the influences of the pulse width and filter bandwidth on the purity, and obtain the optimized condition. The optimized pump pulse width is 8 ps and filter bandwidth is about 40 GHz or less. A home-made 1550.1 nm mode-locked fiber laser source, whose pulse width and repetition rate are 25 ps and 27.9 MHz respectively, acts as a pump of photon source. A tunable attenuator is used to adjust the pump power of the photon source, and the broad band background fluorescence photons are filtered out by cascade 100 GHz DWDM filters. The clean pump beam is divided into two equal parts by the 50 : 50 optical coupler to pump two 300 m DSFs (cooled by liquid nitrogen) to generate independent photon sources. Then the strong pump beam and noise photon from Raman scattering in orthogonal polarization are removed by 2 groups of 200 GHz DWDM filters and fiber polarization rotator and polarizer. Then two 100 GHz DWDMs are used for separating photons at correlated channel pairs. The relative delay between the two independent photons is adjusted by tunable fiber delay line. Photons from the same channels are combined in a second beam splitter for interference, and the other two photons are used as trigger signals. The two triggered photons are detected by two free running InGaAs avalanched single photon detectors (APD1, APD4, ID Quanta, ID220, 20% detection efficiency, 3 s dead time, dark count rate 4k cps), and the outputs of detectors APD1 and APD4 are used to trigger two single-photon detectors running in the gated mode (APD2, APD3, Qasky, Hefei, China, 100 MHz, free gating single photon detectors, 20% detection efficiency, dark count probability 410-5 per gate) for twophoton coincidence measurement. Detection output signals from APD2 and APD3 are sent to our coincidence count device (Pico quanta, TimeHarp 260, 1.6 ns coincidence window) for four-photon coincidence measurement. Before measuring the HOMI, we obtain a maximum-coincidence-to-accidental-coincidence ratio (CAR) of 131 by cooling the fiber in liquid nitrogen when the pump power is 23 W. There are a few remarks we want to point out.Firstly, the photon sources are not operated at the optimized pump pulse width for pure single photon generation, but narrow band 100 GHz filters are used in the experiments to increase the purity of the sources. Secondly, single photon detectors used in our experiment have lower detection efficiency and much higher dark counts than nano-wire single photon detectors, if we have high-performance nano-wire single photon detector, experimental results will be greatly improved due to the four-fold coincidences and dark coincidences scaling quadruplicate with the detection efficiency and dark count probability of a single detector. Thirdly, we use relatively high pump power for each DSF (0.12 mW) to reduce measurement time for photon coincidence, which will lead to a very poor raw visibility certainly. Finally, though only a 100 GHz channel pair is used in our experiment, we can use other channels for multiplexing such interference processes to improve the channel capacity in future quantum communication tasks theoretically. Our study shows greatly promising integrated optical elements for future scalable quantum information processing.