I want to use Giraph as a graph processing tool for my work. I am familiar with Mahout and I know I can use some parts of Mahout such as Recommender System without using Hadoop. However, I don't know if this is the same for Giraph and whether I can use it without using Hadoop.
You need Hadoop in order to run Giraph as the single tasks (master and worker) are internally executed as map-only jobs. Alternatively, you can also run Giraph as a yarn application.
Have a look at the giraph quick start guide and search the mailing list if you get into trouble.