Home ¦ Posts ¦ Books ¦ Articles ¦ Talks ¦ Notes

AWS VPC subnets and Internet connectivity over IPv4

We can have two kinds of subnets inside a AWS VPC - private and public. A public subnet is one which is attached to an Internet Gateway. This essentially adds a routing table entry to the subnet's routing table sending all Internet traffic to an Internet Gateway. On the other hand, if traffic from a subnet destined for the "Internet" is sent to either a NAT instance, or a AWS managed NAT device, the subnet is a private subnet.

An EC2 instance running in either subnet can choose to have a public IP address or not. Once you give your instance a public IP address, it becomes "reachable" from the Internet (let's call it ingress) and of course can reach "Internet" resources from the instance (let's call it egress).

Let's discuss the flow of traffic and the address translations that happens in various cases that may arise in the above scenario.

Public subnet - Public IP

Consider a EC2 instance, E in a public subnet having a public IP.

Ingress

Internet resource, device A talks to our instance E using the public IP. E sees it gets a request from A, responds accordingly. The response goes via Internet Gateway configured for the subnet. The Internet Gateway perfroms a Network Address translation where it changes the source IP address of the response packet to match the public IP address of E, so that A gets the response from E, and not E's internal IP address.

A -> (Public IP) E -> (Routing) -> IGW -> (Change source IP to public IP of E) -> A

Egress

Instance E tries to access an Internet resource, B. The traffic goes via the Internet Gateway where a network address translation takes place - the source IP address is changed from the internal IP to the public facing IP. When the response is received by the Internet Gateway, the destination IP is changed from the public IP of instance E to the private IP.

E -> (Routing) -> IGW -> (Change source IP to public IP of E) -> B -> IGW -> (Change destination IP to private IP of E) -> E

Public subnet - No Public IP

Considering the above address translation, AWS will not prevent you from having an instance with no public IP, but you will not be able to talk to any "Internet" resource and of course, vice-versa - i.e. no ingress or egress from/to Internet resource(s).

Private subnet - Public IP

Consider a EC2 instance, F in a private subnet having a public IP.

Ingress

Internet resource, device B can talk to instance, F using the public IP. F gets the connection request, sends back an acknowledgement. This acknowledgement needs to go to B via either a NAT instance or a AWS managed NAT device. The NAT device performs a Source address translation replacing the private IP of the instance, F by the NAT's public IP address. The acknowledgement makes it way back to the device B's "edge" device (firewall for example) and doesn't recognize the public NAT IP as part of an ongoing connection and just drops it. Hence, although, device B can attempt to initiate connection to F, it never actually succeeds.

A -> (Public IP) F -> (Routing) -> NAT -> (Change source IP to public IP of NAT) -> Drops packet

Egress

F tries to talk an Internet resource, I. The routing table forwards the request to the NAT instance/device. This device then replaces the source IP of the request which currently is the private IP by the NAT device's Public IP addres and forwards the request to the destination resource. When the response is received, the NAT device changes the destination IP address from itself to that of the private IP address of the instance.

F -> (Routing) -> NAT -> (Change source IP to public IP of NAT) -> I -> NAT -> (Change dest IP to private IP of F) -> F

Private subnet - No Public IP

Consider a EC2 instance, G in a private subnet having no public IP

Ingress

In this case, there is no Ingress traffic to G as there is no public IP.

Egress

G tries to talk an Internet resource, J. The routing table forwards the request to the NAT instance/device. This device then replaces the source IP of the request which currently is the private IP by the NAT device's Public IP addres and forwards the request to the destination resource. When the response is received, the NAT device changes the destination IP address from itself to that of the private IP address of the instance.

G -> (Routing) -> NAT -> (Change source IP to public IP of NAT) -> J -> NAT -> (Change dest IP to private IP of G) -> G

Learn more

Share on: Diaspora*TwitterFacebookGoogle+Email

© Amit Saha. Built using Pelican. Customised theme based on the one by Giulio Fidente on github.