The Transport Layer
Table of Contents
- The Transport Layer
- Dissection of a TCP Segment
- TCP Control Flags and the Three-way Handshake
- TCP Socket States
- Connection-oriented and Connectionless-protocols
- System Ports versus Ephemeral Ports
- Firewalls
The Transport Layer
The transport layer is responsible for lots of important functions of reliable computer networking. These include multiplexing and demultiplexing traffic, establishing long running connections and ensuring data integrity through error checking and data verification. By the end of this lesson you should be able to describe what multiplexing and demultiplexing are, and how they work. You'll be able to identify the differences between TCP and UDP, explain the three way handshake, and understand how TCP flags are used in this process. Finally, you'll be able to describe the basics of how firewalls keep networks safe. The transport layer has the ability to multiplex and demultiplex, which sets this layer apart from all others. Multiplexing in the transport layer means that nodes on the network have the ability to direct traffic toward many different receiving services. Demultiplexing is the same concept, just at the receiving end, it's taking traffic that's all aimed at the same node and delivering it to the proper receiving service.
The transport layer handles multiplexing and demultiplexing through ports.
A port is a 16-bit number that's used to direct traffic to specific services running on a networked computer.
Remember the concept of server and clients? A server or service is a program running on a computer waiting to be asked for data. A client is another program that is requesting this data.
Different network services run while listening on specific ports for incoming requests. For example, the traditional port for HTTP or unencrypted web traffic is port 80. If we want tor request the webpage from a web server running on a computer listening on IP 10.1.1.100, the traffic would be directed to port 80 on that computer.
Ports are normally denoted with a colon after the IP address. So the full IP and port in this scenario could be described as 10.1.1.100:80. When written this way, it's known as a socket address or socket number.
The same device might also be running an FTP or file transfer protocol server. FTP is an older method used for transferring files from one computer to another, but you still see it in use today.
FTP traditionally listens on port 21, so if you wanted to establish a connection to an FTP server running on the same IP that our example web server was running on, you direct traffic to 10.1.1.100 port 21. You might find yourself working in IT support at a small business. In these environments, a single server could host almost all of the applications needed to run a business. The same computer might host an internal website, the mail server for the company, file server for sharing files, a print server for sharing network printers, pretty much anything. This is all possible because of multiplexing and demultiplexing, and the addition of ports to our addressing scheme.
Dissection of a TCP Segment
Just like how an Ethernet frame encapsulates an IP datagram, an IP datagram encapsulates a TCP segment. Remember that an Ethernet frame has a payload section which is really just the entire contents of an IP datagram. Remember also that an IP datagram has a payload section and this is made up of what's known as a TCP segment.
A TCP segment is made up of a TCP header and a data section.
This data section, as you might guess, is just another payload area for where the application layer places its data. A TCP header itself is split into lots of fields containing lots of information.
Destination port: the port of the service the traffic is intended for.
Source port: a high numbered port chosen from a special section of ports known as ephemeral ports.
It's enough to know that a source port is required to keep lots of outgoing connections separate. You know how a destination port, say port 80, is needed to make sure traffic reaches a web server running on a certain IP? Similarly, a source port is needed so that when the web server replies, the computer making the original request can send this data to the program that was actually requesting it. It is in this way that when it web server responds to your requests to view a webpage that this response gets received by your web browser and not your word processor.
Sequence Number: This is a 32-bit number that's used to keep track of where in a sequence of TCP segments this one is expected to be.
You might remember that lower on our protocol stack, there are limits to the total size of what we send across the wire. In Ethernet frame, it's usually limited in size to 1,518 bytes(Refers to Maximum_transmission_unit), but we usually need to send way more data than that. At the transport layer, TCP splits all of this data up into many segments. The sequence number in a header is used to keep track of which segment out of many this particular segment might be.
Acknowledgment Number:the number of the next expected segment.
In very simple language, a sequence number of one and an acknowledgement number of two could be read as this is segment one, expect segment two next.
Data offset field:a four-bit number that communicates how long the TCP header for this segment is.This is so that the receiving network device understands where the actual data payload begins.
TCP window:specifies the range of sequence numbers that might be sent before an acknowledgement is required.As we'll cover in more details soon, TCP is a protocol that's super reliant on acknowledgements. This is done in order to make sure that all expected data is actually being received and that the sending device doesn't waste time sending data that isn't being received.
TCP checksum:operates just like the checksum fields at the IP and Ethernet level.Once all of this segment has been ingested by a recipient, the checksum is calculated across the entire segment and is compared with the checksum in the header to make sure that there was no data lost or corrupted along the way.
Urgent pointer field:used in conjunction with one of the TCP control flags to point out particular segments that might be more important than others.This is a feature of TCP that hasn't really ever seen adoption and you'll probably never find it in modern networking.
Options field:Like the urgent pointer field, this is rarely used in the real world, it's sometimes used for more complicated flow control protocols.
Padding: just a sequence of zeros to ensure that the data payload section begins at the expected location.
TCP Control Flags and the Three-way Handshake
As a protocol, TCP establishes connections used to send long chains of segments of data. You can contrast this with the protocols that are lower in the networking model. These include IP and Ethernet, which just send individual packets of data.You need to understand exactly how that works, so you can troubleshoot issues, where network traffic may not be behaving in the expected manner. The way TCP establishes a connection, is through the use of different TCP control flags, used in a very specific order.
Before we cover how connections are established and closed, let's first define the six TCP control flags. We'll look at them in the order that they appear in a TCP header.
- URG(Urgent) :
A value of one here indicates that the segment is considered urgent and that the urgent pointer field has more data about this. Like we mentioned in the before, this feature of TCP has never really had wide spreaded option and isn't normally seen. - ACK(Acknowledge) :
A value of one in this field means that the acknowledgment number field should be examined. - PSH(Push) :
The transmitting device wants the receiving device to push currently- buffered data to the application on the receiving end as soon as possible.
A buffer is a computing technique, where a certain amount of data is held somewhere, before being sent somewhere else. This has lots of practical applications. In terms of TCP, it's used to send large chunks of data more efficiently. By keeping some amount of data in a buffer, TCP can deliver more meaningful chunks of data to the program waiting for it. But in some cases, you might be sending a very small amount of information, that you need the listening program to respond to immediately. This is what the push flag does. - RST(Reset) :
One of the sides in a TCP connection hasn't been able to properly recover from a series of missing or malformed segments. It's a way for one of the partners in a TCP connection to basically say, "Wait, I can't put together what you mean, let's start over from scratch." - SYN(Synchronize) :
It's used when first establishing a TCP connection and make sure the receiving end knows to examine the sequence number field. - FIN(Finish) :
When this flag is set to one, it means the transmitting computer doesn't have any more data to send and the connection can be closed.
For a good example of how TCP control flags are used, let's check out how a TCP connection is established. Computer A will be our transmitting computer and computer B will be our receiving computer. To start the process off, computer A, sends a TCP segment to computer B with this SYN flag set. This is computer A's way of saying, "Let's establish a connection and look at my sequence number field, so we know where this conversation starts." Computer B then responds with a TCP segment, where both the SYN and ACK flags are set. This is computer B's way of saying, "Sure, let's establish a connection and I acknowledge your sequence number." Then computer A responds again with just the ACK flag set, which is just saying, "I acknowledge your acknowledgement. Let's start sending data." I love how polite they are to each other. This exchange involving segments that have SYN, SYN/ACK and ACK sets, happens every single time a TCP connection is established anywhere. And is so famous that it has a nickname. The three way handshake.
A handshake is a way for two devices to ensure that they're speaking the same protocol and will be able to understand each other.
Once the three way handshake is complete, the TCP connection is established. Now, computer A is free to send whatever data it wants to computer B and vice versa. Since both sides have now sent SYN/ACK pairs to each other, a TCP connection in this state is operating in full duplex. Each segment sent in either direction should be responded to by TCP segment with the ACK field set. This way, the other side always knows what has been received.
Once one of the devices involved with the TCP connection is ready to close the connection, something known as a four way handshake happens. The computer ready to close the connection, sends a FIN flag, which the other computer acknowledges with an ACK flag. Then, if this computer is also ready to close the connection, which will almost always be the case. It will send a FIN flag. This is again responded to by an ACK flag. Hypothetically, a TCP connection can stay open in simplex mode with only one side closing the connection. But this isn't something you'll run into very often.
TCP Socket States
A socket is the instantiation of an endpoint in a potential TCP connection. An instantiation is the actual implementation of something defined elsewhere.
TCP sockets require actual programs to instantiate them. You can contrast this with a port which is more of a virtual descriptive thing. In other words, you can send traffic to any port you want, but you're only going to get a response if a program has opened a socket on that port. TCP sockets can exist in lots of states. And being able to understand what those mean will help you troubleshoot network connectivity issues.We'll cover the most common ones here.
- LISTEN
A TCP socket is ready and listening for incoming connections.You'd see this on the server side only. - SYN_SENT
A synchronization request has been sent, but the connection hasn't been established yet. You'd see this on the client side only. - SYN_RECEIVED
A socket previously in a listener state, has received a synchronization request and sent a SYN_ACK back. But it hasn't received the final ACK from the client yet. You'd see this on the server side only. - ESTABLISHED
The TCP connection is in working order, and both sides are free to send each other data.You'd see this state on both the client and server sides of the connection. This will be true of all the following socket states, too. So keep that in mind. - FIN_WAIT
A FIN has been sent, but the corresponding ACK from the other end hasn't been received yet. - CLOSE_WAIT
The connection has been closed at the TCP layer, but that the application that opened the socket hasn't released its hold on the socket yet. - CLOSED
The connection has been fully terminated, and that no further communication is possible.
There are other TCP socket states that exist. Additionally, socket states and their names, can vary from operating system to operating system. That's because they exist outside of the scope of the definition of TCP itself. TCP, as a protocol, is universal in how it's used since every device speaking to TCP protocol has to do this in the exact same way for communications to be successful. Choosing how to describe the states of a socket at the operating system level isn't quite as universal. When troubleshooting issues at the TCP layer, make sure you check out the exact socket state definitions for the systems you're working with.
Connection-oriented and Connectionless-protocols
So far, we've mostly focused on TCP, which is a connection-oriented protocol.
A connection-oriented protocol is one that establishes a connection, and uses this to ensure that all data has been properly transmitted.
A connection at the transport layer implies that every segment of data sent is acknowledged. This way, both ends of the connection always know which bits of data have definitely been delivered to the other side and which haven't. Connection-oriented protocols are important because the Internet is a vast and busy place. And lots of things could go wrong while trying to get data from point A to point B. You might remember from our lesson about the physical layer that even some minor crosstalk from a neighboring twisted pair in the same cable can be enough to make a cyclical redundancy check fail. This could cause the entire frame to be discarded, yikes. If even a single bit doesn't get transmitted properly, the resulting data is often incomprehensible by the receiving end. And remember that at the lowest level, a bit is just an electrical signal within a certain voltage range.
But there are plenty of other reasons why traffic might not reach its destination beyond line errors. It could be anything. Pure congestion might cause a router to drop your traffic in favor of forwarding more important traffic. Or a construction company could cut a fiber cable connecting two ISPs, anything's possible. Connection-oriented protocols, like TCP, protect against this by forming connections and through the constant stream of acknowledgments. Our protocols at lower levels of our network model, like IP and Ethernet, do use checksums to ensure that all the data they received was correct. But did you notice that we never discussed any attempts at resending data that doesn't pass this check? That's because that's entirely up to the transport layer protocol.
At the IP or Ethernet level, if a checksum doesn't compute, all of that data is just discarded. It's up to TCP to determine when to resend this data. Since TCP expects an ack for every bit of data it sends, it's in the best position to know what data successfully got delivered and can make the decision to resend a segment if needed. This is another reason why sequence numbers are so important. While TCP will generally send all segments in sequential order, they may not always arrive in that order. If some of the segments had to be resent due to errors at lower layers, it doesn't matter if they arrive slightly out of order. This is because sequence numbers allow for all of the data to be put back together in the right order. It's pretty handy. Now, as you might have picked up on, there's a lot of overhead with connection-oriented protocols like TCP. You have to establish the connection. You have to send a stream of constant streams of acknowledgements. You have to tear the connection down at the end. That all accounts for a lot of extra traffic. While this is important traffic, it's really only useful if you absolutely, positively have to be sure your data reaches its destination.
You can contrast this with connectionless protocols. The most common of these is known as UDP, or User Datagram Protocol. Unlike TCP, UDP doesn't rely on connections, and it doesn't even support the concept of an acknowledgement. With UDP, you just set a destination port and send the packet. This is useful for messages that aren't super important. A great example of UDP is streaming video. Let's imagine that each UDP datagram is a single frame of a video. For the best viewing experience, you might hope that every single frame makes it to the viewer, but it doesn't really matter if a few get lost along the way. A video will still be pretty watchable unless it's missing a lot of its frames. By getting rid of all the overhead of TCP, you might actually be able to send higher quality video with UDP. That's because you'll be saving more of the available bandwidth for actual data transfer, instead of the overhead of establishing connections and acknowledging delivered data segments.
System Ports versus Ephemeral Ports
Transportation layer protocols use a concept of ports and multiplexing/demultiplexing to deliver data to individual services listening on network nodes. These ports are represented by a single 16-bit number, meaning that they can represent the numbers 0-65535.
This range has been split up by the IANA (Internet Assigned Numbers Authority) into independent sections:
Port 0 isn’t in use for network traffic, but it’s sometimes used in communications taking place between different programs on the same computer.
Ports 1-1023 are referred to as system ports, or sometimes as “well-known ports.” These ports represent the official ports for most well-known network services. In an earlier video, we talked about how HTTP normally communicates over port 80, while FTP usually communicates over port 21. In most operating systems, administrator-level access is needed to start a program that listens on a system port.
Ports 1024-49151 are known as registered ports. These ports are used for lots of other network services that might not be quite as common as the ones that are on system ports. A good example of a registered port is 3306, which is the port that many databases listen on. Registered ports are sometimes officially registered and acknowledged by the IANA, but not always. On most operating systems, any user of any access level can start a program listening on a registered port.
Finally, we have ports 49152-65535. These are known as private or ephemeral ports. Ephemeral ports can’t be registered with the IANA and are generally used for establishing outbound connections. You should remember that all TCP traffic uses a destination port and a source port. When a client wants to communicate with a server, the client will be assigned an ephemeral port to be used for just that one connection, while the server listens on a static system or registered port.
Not all operating systems follow the ephemeral port recommendations of the IANA. In this lesson, we’ll continue to assume that the ephemeral ports used for outbound connections consist of the ports 49152 through 65535. But it’s important to know that this exact range can vary depending on the platform you’re working on. Sometimes portions of the registered ports range are used, but no modern operating system will ever use a system port for outbound communication
To learn more about ports, and to see a list of what ports have been assigned to what services, check out the IANA Service Name and Transport Protocol Port Number Registry. A similar list on Wikipedia is not official, but it is a little easier to read. Check it out, too!
Firewalls
A firewall is just a device that blocks traffic that meets certain criteria.
Firewalls are a critical concept to keeping a network secure since they are the primary way you can stop traffic you don't want from entering a network.
Firewalls can actually operate at lots of different layers of the network. There are firewalls that can perform inspection of application layer traffic, and firewalls that primarily deal with blocking ranges of IP addresses. The reason we cover firewalls here is that they're most commonly used at the transportation layer.
Firewalls that operate at the transportation layer will generally have a configuration that enables them to block traffic to certain ports while allowing traffic to other ports. Let's imagine a simple small business network. The small business might have one server which hosts multiple network services. This server might have a web server that hosts the company's website, while also serving as the file server for a confidential internal document.
A firewall placed at the perimeter of the network could be configured to allow anyone to send traffic to port 80 in order to view the web page. At the same time, it could block all access for external IPs to any other port. So that no one outside of the local area network could access the file server.
Firewalls are sometimes independent network devices, but it's really better to think of them as a program that can run anywhere. For many companies and almost all home users, the functionality of a router and a firewall is performed by the same device. And firewalls can run on individual hosts instead of being a network device. All major modern operating systems have firewall functionality built-in. That way, blocking or allowing traffic to various ports and therefore to specific services can be performed at the host level as well.
References:
https://www.coursera.org/learn/computer-networking/lecture/cboIM/the-transport-layer
https://www.coursera.org/learn/computer-networking/lecture/EYfgW/dissection-of-a-tcp-segment
https://en.wikipedia.org/wiki/Transmission_Control_Protocol
https://www.coursera.org/learn/computer-networking/supplement/GJHb4/supplemental-reading-for-system-ports-versus-ephemeral-ports
https://www.coursera.org/learn/computer-networking/lecture/hGnHm/tcp-control-flags-and-the-three-way-handshake
https://www.coursera.org/learn/computer-networking/lecture/1ELOr/tcp-socket-states
https://www.coursera.org/learn/computer-networking/lecture/mlUNd/connection-oriented-and-connectionless-protocols
https://www.coursera.org/learn/computer-networking/lecture/7v4n0/firewalls