Sockets

What are they? 🔌

Ye Lin Aung

What is the first thing that comes to mind?

This is about (Linux) sockets

• It is a way to send messages across a network
• Imagine Bob want to say something to Alice
• Bob has a number “111-111” and Alice has “111-222”
• Bob can call Alice only when Bob knows the Alice’s number.
• Alice can accept or reject the incoming call.
• Upon the acceptance from Alice, Bob can send some messages.
• Likewise, Alice can return some messages.

What type of Sockets are there?

• Unix Sockets for Inter Process Communication (IPC)
• Sending messages back and forth between “processes” in the same OS
• Look for “Docker socket” for example!
• Network Sockets (IP Sockets)
• TCP Sockets
• UDP Sockets

What’s below and above them ?

How does the TCP/UDP sockets work?

• Imagine Bob is the “Client”
• Alice is the “Server”
• This is a TCP Socket
• Also known as Berkeley Socket

How are they sending the data over ?

• Over the internet, the data is going through many devices
• Routers
• Switches
• Hub

How are they sending the data over ?

• Locally, (i.e within your own computer)
• There is a special network device called “loopback” interface
• Also known as 127.0.0.1
• Which is bind to “localhost” in /etc/hosts

Demo time!

The (in)famous TCP 3-Way Handshake 🤝

SYN – SYN,ACK – ACK 🤝

SYN – SYN,ACK – ACK 🤝 but in Terminal

“pdb” ( 🐍 Python Debugger) cheatsheet

• b: set a breakpoint
• c: continue debugging until you hit a breakpoint
• s: step through the code
• n: to go to next line of code
• l: list source code for the current file (default: 11 lines including the line being executed)
• u: navigate up a stack frame
• d: navigate down a stack frame
• p: to print the value of an expression in the current context

This is cool but … 🤔

• That “toy” server dies after doing some stuff with an incoming request ???
• Maybe one request per process is the what they cool kids call “Lambda” ??
• What if we want to accept multiple connections at the same time?
• And process them “asynchronously”
• Can we have the 🎂 and eat it too ???
• Hmmm….🤔

Enter “concurrency” 🥁

• There are many ways to do it - in Python
• The popular "asyncio" — Asynchronous I/O
• Good ol’ threading
• But … GIL be watching you
• Or .. we go back to one request, one process but in a cooler way with multiprocessing — Process-based parallelism 😉

But today .. we are going back to stone age

• Here comes “selectors”
• selectors — High-level I/O multiplexing
• After all, network sockets are nothing more than “input/output” (IO) operations
• The more efficient operations can be, the more requests we can proces
• I didn’t say faster lol
• There are many iteration of this throughout.
• In Linux/Unix-like systems, there are select(), poll(), epoll()

OK enough with all these slides! Show me the code

• Here are the scripts I used in the demo

https://gist.github.com/yelinaung/57e817eeac0c89487a5a3a621ee6eb69

I/O Multiplexing

• A technique that allows a program to monitor multiple input/output (I/O) channels simultaneously (sockets)
• Without the need for multiple threads or processes
• Like a juggler who can keep an eye on several balls in the air at once, rather than focusing on just one ball at a time!

What are select(), poll(), epoll() ?

• There are “system call” (an interface for User/Kernel space) in Linux
• They syscalls allow a process to monitor multiple socket to read, write, or handle exceptions without blocking on each socket individually.
• There are small differences and similarities between each of them but basically, in terms of performance and scalability, it is like

​

epoll() > poll() > select()

There’s a new kid in town “io_uring"

• A new “IO” interface in Linux
• https://unixism.net/loti/what_is_io_uring.html

A bonus - fetch( .. )

• Browser API that fetches things given a valid URL
• Fetch API
• Yes, it’s based on the same concept
• But .. there are much more going-on …

What’s next?

• Ok, TCP was cool. Where do I learn more about it ?
• Building Blocks of TCP - High Performance Browser Networking
• How about UDP ? Does it do the same “3 way handshake” ?
• (Hint: no)
• How about TLS ?
• How about Async ? Eventloop ? Maybe, that’s the next talk!
• Finally, what is the point of all these?

Thank you

Link to the talk https://bit.ly/yelin-socket

Feedback pls 🙏

https://bit.ly/yelin-talk-1-feedback