Week 0 - Assumed knowledge
Notes from Fundamentals course
I followed this Fundamentals course to understand the basics.
Lecture 1: Hosts and devices
Some terminology first:
These terms are relative to a single request a host can be a client in another request.
Signal strength decays over long distances.
When joining lots of computers together there is a scaling problem if they all need to connect to one another. So instead we us a single entity to act as an intermediary.
Routers make the hierarchy on the internet possible by owning the prefix for that subnet.
Whilst these devices are called routers and switches there is a more abstract concept here.
- Switching: The process of moving data within networks.
- Routing: The process of moving data between networks.
Many other devices can perform switching and routing other than a switch and a router.
Lesson 2: Open Systems interconnection (OSI) model
Open Systems interconnection (OSI) model
Difference between an IP and MAC address
The main other model of the internet is IPS model
Connection between OSI and IPS models
Lesson 3: How do hosts talk to one another
There are two important different cases to cover.
- How host talk to one another on the same network, and
- How hosts talk to one another on different networks.
Talking on the same network
When hosts talk to each other on the same network they need 3 identifying bits of information.
- MAC address,
- Internet Protocol (IPv4), and
- Network mask (this tells the host which other hosts are on its network).
Suppose we need host A to send data to host B.
As host A is already part of the network it has the network mask already.
We will assume host A already has host B’s Internet Protocol (IPv4) through a DN entry or an act of god.
The only further information about host B host A needs is its MAC address. To do this it will need to use the Address Resolution Protocol (ARP) as discussed before.
Every host has an ARP cache that stores the MAC address against each Internet Protocol (IPv4) it knows in the network. If host B’s address is already in host A’s ARP cache it can use that to generate the layer 2 and 3 headers.
If not it will need to make an ARP request to the network. It broadcasts its IP and MAC address on the network asking for the host at host B’s IP address to respond. It does this by setting in the layer 2 header the all f’s MAC address - an address reserved for the purpose of broadcasting to the local network.
Once host B gets this ARP request it can populate its ARP cache entry for host A then send an ARP response directly back (Unicast) to host A with its MAC address.
Now with host B’s MAC address host A can now safely send the data directly to host B.
This is how computers on the same network communicate with each they use ARP to populate their ARP cache so they can directly communicate with one another.
Talking to computers on different networks
In this set up host A needs to send data to host C but there is a router in-between them. We assume host A already has host C’s IP address.
It will know this IP is on a foreign network using host C’s IP address and its own sub-net mask.
The first step for host A to send data to host C is to get it to its local router. It will know the address of that router as it will be set to the default gateway for host A. If this is the first time host A has connected to the router it can use ARP to find the routers IP address.
Host A now constructs the frame to send to the router it sets the layer 3 header using its IP address and host C’s IP address. The it sets the layer 2 header using its MAC address and the routers MAC address and sends it off unicast to the network.
At this point the router takes the frame it receives strips out the layer 2 header to get a packet then it uses it’s own ARP table to attach a new layer 2 header to send it further on. This could be directly to host C or to another router.
Lesson 4: Everything switches do to communicate within a network
Rules of a switch
These are the rules for Switching so applies to anything that can do switching.
When performing switching they only care about the layer 2 data so they don’t know anything about IP addresses - only MAC addresses.
In a switch each device connects to a different port. The switch will maintain a MAC address table that maps different ports to MAC addresses. There are then 3 operations a switch can do.
- Learn,
- Flood, and
- Forward.
When a switch gets a frame from a new host it can learn that hosts MAC address as it will be the source MAC address of that frame. So it can add this to its MAC address table.
If the switch gets a message for a MAC address it doesn’t know it will flood the network with that frame to make sure it gets to the intended host. This involves duplicating the frame and sending it on all ports that wasn’t the port the package came from.
The idea is that hosts that do not match that MAC address throw away the frame as it is not for them.
If the switch already knows the hosts port in its MAC address table it forwards that frame on to the associated port for the destination.
As this all works on layer 2 and doesn’t involve IP addresses the router on the network is just another host from the switches perspective.
Traffic going through a switch does not use a switches MAC address or IP address. However switches are hosts on that network so have an IP address and MAC address. These only get used if you are looking to connect directly to the switch to configure it.
Unicast flooding vs Brodcasts
Flooding is an action the switch takes when it does not know the port for a provided MAC address. This replicates the frame and sends it on all available ports.
Broadcast is a type of frame that has the destination MAC address set as all f’s. These will always get flooded by the switch as it will not have a port for the MAC address of all f’s.
Virtual Local Area Networks (VLAN)
Multiple switches
To handle loops in the network we have to use the Spanning Tree Protocol (STP) to turn off certain connections.
Lesson 5: Everything a router does to communicate between networks
In RFC2460 Internet Protocol, Version 6 (IPv6) Specification it defines what a node, host, and router is.
In other words the only difference between a router and a host is that a host will drop a packet not matching its own IP address whereas a router will do its best to get that packet to its home.
A router keeps a table of all networks it knows about in its routing table. Different networks are identified by the part of the IP space associated to that network. The router will bind these different network spaces to interfaces of the router.
If a router gets a packet it does not know the address of it will just drop that packet.
There are many different Dynamic routing protocols RIP, OSPF, BGP, EIGRP, IS-IS, ect They differ in how they discover new routers, what addresses they share, and how they share thous addresses.
Difference between an ARP table and a Router table
As routers have IP addresses on the networks associated to their interfaces they therefor have to have ARP tables for these networks.
The main difference between ARP tables and router tables is that ARP tables always start empty and can be fully populated through ARP. If the router gets a frame it doesn’t have an ARP entry for it will use ARP to discover it - whereas if a router gets a packet it doesn’t have a routing entry for it will drop that packet.
Hierarchy helps with scaling
If you use a tree like network to connect routers the longest path grows like $\log(n)$ instead of linear designs that will grown as $n$. However there are more reasons to use a Hierarchy a base on the IP address.
With the entries in the routing table the subnet mask indicates what part of the IP address you need to match on. For example the routing table may look something like
| IP address | Subnet mask | Interface |
|---|---|---|
| 10.40.55.0 | 24 | R1 |
| 10.20.0.0 | 16 | R2 |
the first entry would say any IP address matching the first 24 bits or the first 3 decimals of the IP address go here whereas the second entry says any address matching the first 16 bits or 2 decimals then go here. The second entry is called route summarization. It is actually referring to a lot of subnets but groups them all index one entry as the first step is always to go to a single router. This can only be done if you arrange your IP addresses in a hierarchical manner.
You can set a default route - this is one that matches on 0 bits. Then instead of dropping any unknown address you instead send it to the default route.
If you have a default gateway then it will match with all addresses which might cause a conflict. By default the more specific a routing entry the higher priority that rule. Though rules can get more complicated than this.
Lesson 6: Protocols
Protocol: Set of rules and messages that form an Internet standard.
Some were already described above. Such as ARP (RFC 826).
Simple Mail Transfer Protocol (SMTP)
Hyper Text Transfer Protocol (HTTP)
Transport Layer Security (TLS)
Dynamic Host Configuration Protocol (DHCP)
Preparation questions
- What is a “protocol”? What are some of the most well-known and used protocols?
A protocol is a set of messages and rules that are an internet standard. They can be found defined in RFC’s. Some popular protocols are ARP, HTTP, TLS, and DNS.
- What is an ISP? What is an AS?
Internet Service Provider (ISP)
- What is the OSI layer model? What is the primary responsibility of each layer?
The Open System Interconnection model is the 7 conceptual layers of how connected systems work.
- Physical: Cables and transportation of data.
- Data Link: Devices that move data on and off the physical layer. Identifies actors by MAC addresses.
- Network: Devices that work as the ultimate destination for data within a network. Uses IP addresses to distinguish actors.
- Transport: The passes data going to the same host to different processes. This uses ports to distinguish entities.
- Session: This allows users using the same process to be distinguished.
- Presentation: This determines how the data should be interpreted.
- Application: How the data will be used to run commands.
- How does the layered architecture of the internet allow fundamentally different technologies (such as WiFi vs. Ethernet) to be used together?
It established clear separation of concerns and the interfaces in which the operate by. For example for ethernet or Wifi as long as the data has been encapsulated properly these Physical layers are only responsible for getting data to the next entity in the network.
- What is the client-server model? What is the peer-to-peer model? What are the strengths and weaknesses of each?
- What is a port (number)? How is it used?
A port number is how Layer 4 distinguishes different applications. It is used to separate up messages going to the same host.
- What’s the difference between a well-known port number and an ephemeral port?
- What is TCP? What is UDP? What are some of the major differences between them?
Transmission Control Protocol (TCP)
- If you were developing a brand new application layer protocol, what aspects of TCP would make it more appealing to use? What aspects of UDP would make it more appealing to use?
I would use TCP if I wanted to gaurentee integrity of the data and didn’t care about latency whereas I would use UDP if I wanted a fast connection where packet replay was not important.
- What is a socket? How is it used?
- Can you generally describe how HTTP works in an example (how the HTTP request is initiated, how the webserver receives the request, how the content is returned, etc.)?
Yes I think so but that is a lot of writing.
- What is an IP address? How is it different than a MAC Address?
An IP address is the unique identifier for a host. This works for hosts that are not not the same network. This will be added to the message at Layer 3 and is responsible for checking the message is going to the correct host. MAC addresses are encapsulated at layer 2 and are only responsible for getting the frame to the correct host within a network.
- What is a switch? What is a router? What are the major differences between the roles that they play in a network, classically?
Switches are responsible for getting data between hosts on the same network. The have an ARP table and can do 3 operations, learn, flood and forward. These allow for the switch to learn about new hosts that are connected to its ports and find their MAC addresses. This device only cares about the layer 2 information of the message.
Routing on the other hand is responsible for getting packets to hosts that are not on their network. The maintain a route table that informs them about the IP addresses it knows where to route packets to. This operates on the Layer 3 information of the packet.
- What is a default gateway
The default gateway is the address of the main router in that network it is IP address that hosts add the MAC address of if the IP address is not in this network.
- What happens to a message as it gets sent across a network, as far as encapsulation and protocol headers?
Suppose we have a message we need to send to an application first we encode that information using the specification of layer 6 of the target application. We add headers to identify this user session in layer 5. Then we add the randomly assigned source port of the application sending the data and the destination port of the application we need to send the information to. Now we add the IP address of the machine sending the data and the IP address of the machine we want to send the data to. Then we use that IP address and the subnet mask to determine if the machine is on our network or a foriegn network. Then we either uses its address or the default gateway’s address to look a MAC address of the next hop in the network in our ARP table - if we don’t have it we use ARP to determine that MAC address in the network.
- What is DNS? At a high level, how does it work?
Domain Name Service map domain names to IP addresses. When connecting to a domain name like www.example.com first you go to your domain name provider to look up the associated IP address then once you have done that you can send the message to the server. Note this doesn’t resolve the port of the application this would need to be known before hand. Lots of standard protocols use default ports.