1. What this is about:

• Created a 3-server distributed hash table model using C++ and the necessary socket programming libraries
• Client-Server communications were conducted through bidirectional UDP sockets
• Server-Server communications were conducted through TCP sockets
• Ports were dynamically obtained on the client side
• Servers & Clients are run as separate processes on nunki.usc.edu
• Clients request information by submitting a search query to Server 1
• The Servers act in a recursive manner to obtain the corresponding value and return it to the client
• Note: All IPs have been retrieved using 'localhost' as the host name

2. File descriptions

• client1.cpp

  • Creates and fills a map that stores key-value pairs based on the information in client1.txt
  • Queries the user for a search command and matches it to a key within the map
  • Retreives address information of Server 1 (well-known port number 21564)
  • Creates UDP socket and submits GET request (of the key) to the server using the necessary functions
  • Recieves a POST message from Server 1 with the associated value
  • Closes the UDP socket and terminates
  • At any time an error is detected, the program will output the location of the error and exit

• client2.cpp

  • Does exactly the same as above except it reads data from client2.txt

• dhtserver1.cpp

  • Gets it's own address information using getaddrinfo("localhost",...)
  • The UDP port number is well-known (static) - 21564
  • Binds socket to port & IP and enters a while loop that will let it communicate with as many clients as necessary
  • Note that Server 1 will only cease to operate when ctrl+C is pressed by the user
  • Inside the while loop, Server 1 will receive the GET message from the Client
  • If it has the requested value, it will respond with a POST message and the corresponding value
  • If not, it will call a function that sets up a TCP connection with Server 2
  • In this function, it will obtain Server 2's address info (although the port number is well-known: 22564)
  • Then, a new TCP socket is created and bound to Server 1. It will forward the GET request through this socket
  • After receiving the POST value, it will return to the main function and forward the POST message to the Client
  • Before returning, it will close the TCP socket
  • At any time an error is detected, the program will output the location of the error and exit

• dhtserver2.cpp

  • Creates a TCP socket with well-known port number 22564 and listens for connection requests
  • In a while loop, it accepts all incoming connections and a new socket is created
  • Translates the GET request and responds with a POST message or calls a function to setup a TCP connection with Server 3 correspondingly
  • The function operates exactly like the one in Server 1
  • Once the TCP connection with Server 3 is terminated, the POST message is returned to Server 1
  • At any time an error is detected, the program will output the location of the error and exit

• dhtserver3.cpp

  • Creates a TCP socket with well-known port number 23564 and listens for connection requests
  • In a while loop, it accepts all incoming connections and a new socket is created
  • Receives the GET request and responds with the corresponding POST value
  • At any time an error is detected, the program will output the location of the error and exit

3. How to Run

• A Makefile is included. This will compile all of the separate files of code
• 'make clean' will remove all the executables previously created
• In three separate windows, please run ./server1, ./server2 and ./server3
• In a fourth window, ./client1 may be run
• Client 1 will terminate when its job is done
• In the same window, ./client2 can be run now
• Note: At the end, all the Servers must be terminated with ctrl+C

4. Format of Exchanged Messages

• The messages were sent and received between Clients and Servers as char*
• Many of the operations within the client/server were performed by converting these char*s to strings

5. Possible Idiosyncrasies

• I am unsure of whether this is an idiosyncrasy or not but - My servers and clients bind to the dynamically obtained TCP/UDP port numbers. If I do not bind, the port numbers remain 0 on one side but are accessible on the other (in a client-server scenario). One of the websites from the hints.pdf (https://www.cs.rutgers.edu/~pxk/417/notes/) encourages this to allow for the identification of the socket (specifically section 2 in - https://www.cs.rutgers.edu/~pxk/417/notes/sockets/index.html).
  • NOTE 1: The program works perfectly fine without the bind calls BESIDES printing out the correct port number on one side.
  • NOTE 2: I would like to clarify that the port numbers are certainly dynamically obtained.
  • I hope this isn't too much of a problem
• If no value is found, the receieve function in Server 2 fails since the error message causes the buffer to overflow but the project document said that we do not need to account for this edge case

6. Reused Code

• I have clearly identified where I have reused code in the project
• The main areas are
  • The getaddrinfo() sections have a few (2-4) lines from Beej's Socket Programming Guide
  • The bind() sections have a few (2-4) lines from this website: https://www.cs.rutgers.edu/~pxk/417/notes/sockets/index.html
  • These lines are basically information to preload a sockaddr with