1. Yes you can run JanusGraph on windows. I haven't personally run it on windows and the only real advice I would have is make sure you're running java 8.
2. If you choose Cassandra as your backend you should run at least 3 nodes. If you choose to use elasticsearch to enable mixed indexes you should also run at least a 3 node cluster. The hardware requirements really depend on your use case. I'm able to run a functional cluster in docker containers locally on my macbook pro for small datasets. You'll need to scale your backend infrastructure accordingly with your data. If you have a lot of concurrent requests you'll need to run multiple instances of JanusGraph to spread out the load.
3. The storage backend is all you absolutely need. Depending on your use case you may want to also run a mixed index backend like elasticsearch, lucene, or solr.
4. To the best of my knowledge currently there is no GUI for JanusGraph. The majority of operations can be handled with the gremlin console CLI. Some advanced operations, like bulk loading data to a remote instance, may require writing a standalone application.
Once you get JanusGraph up and running and import your data something else to be aware of is that to really use JanusGraph well you need to look at the
Tinkerpop documentation as well. The JanusGraph docs do cover a lot of the same info on traversal steps, but aren't complete on their own.
Let me know if you have anymore questions and I'll do my best to help answer them.