Thanks for asking.
I am going to make some assumptions in my response and will list them when relevant. Please let us know if they are incorrect.
Assumption: The cameras are in the same building and/or can be streamed back to a single location.
Question #1 Answer: NVR appliances are easier to use, to the extent that software and storage are preloaded. However, COTS servers allow you to scale to higher camera counts and be more flexible in how and where you store video.
Important, how much cost and flexibility tradeoff exists depends on what NVR / VMS manufacturer you are selecting. Do you have a short list of companies you are considering?
Question #2 Answer: Computing bandwidth depends on many factors including the resolution and frame rate (which you've provided). However, it also depends on compression level, the scene complexity (i.e., is the camera watching an empty hallway or a busy intersection) and differences in how cameras handle video.
The best way to compute bandwidth is to get a model of each camera you plan to use and measure it at your site.
For more, How to Calculate Surveillance Storage / Bandwidth
Question #3 Answer: 64 cameras per COTS server is a common specification. However, how many cameras a server can handle depends on the server specified and what you are doing.
Do you plan to do motion detection on the server side or plan to run any analytics on the server? Either of these will significantly increase demands on the servers and could reduce your camera count.
For more, see Specifying VMS Server Size.
Question #3 Part 2 Answer: For specifications to include for COTS server, I am assuming you mean hardware specs, like CPU, RAM, NIC, etc.
This is something you should ask the VMS / NVR manufacturer after you pick them as it can vary depending on manufacturer.
Assumption: I am assuming you are going to record continuously. If you record on motion only, storage will be reduced typically between 30% and 70%.
Question #4 Answer: How much storage needed depends on bandwidth consumed. For a 1.3MP, 12fps, bandwidth can easily range from 300Kb/s to 6Mb/s (assuming H.264, default compression). That's a huge range. Again, as mentioned above it depends on scene complexity and how the camera handles video compression. See: Advanced Camera Bandwidth Test Results
To get a ballpark, let's say your average camera consumes 2Mb/s, and records continuously for 60 days. That would be ~300 TB.
As for getting a SAN, I think network based storage would provide greater flexibility and expansion, especially since that's a fair amount of storage (100+ hard drives will be needed).
Question #5: What do you mean by workstations? Do you mean PCs for operators to view video?