You may plant two sets of 20 lines (20 inbreds). One set from which you can collect pollen and another set where you will pollinated the inbreds. Number of plant or length of row will depends how many seeds you will need to generate from the crosses.
It is better to sow/plant two sets, only for pollen collection, to make available the pollen for fertilization. In this case sowing/planting pollen parent should be done in two different dates to ensure the pollen availability.
It is little bit laborious as well as difficult to make diallel crosses from 20 inbred lines. If you have plan for half diallel your cross will be 190 and for full diallel it will be 380. Close observation and concentration is required for maintaining the crosses. Otherwise it will be a very nice work.
If you are doing General and Specific Combining Ability (GCA and SCA) experiments for evaluating some traits, you can use the diallele cross scheme, as mentioned by Ashraful. You may also want to throw in a couple of 'elite' inbred lines for comparison. Probably also need a couple of locations to grow them. So, the total cross numbers may increase.
It depends on the type of plant. In any case you should have grown the plants in to two different time for vegetables the period between each time 1.0-1.5 week to make sure you will have enough amount of pollen. When you say how many rows the more plots the easier will be to do your job.
I would like to thank you for your question. You must have 3-4 sowing dates from your genotypes with 15 days intervals between each sowing date and each genotype must sow in 3-4 rows 2 meters long.