Osaka University EE ES Talk series Part 2 of 3 3-JUL-2018

A part of Electrical Engineering Lecture Series 2018 at School of Engineering Science, Osaka University / 大阪大学基礎工学部電気工学特別講義2018 2/3

Transcript

1. oueees-201806 talks Part 2/3:Routing and transport protocols Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 1
2. Kenji Rikitake 3-JUL-2018 School of Engineering Science Osaka University Toyonaka, Osaka, Japan @jj1bdx Copyright ©2018 Kenji Rikitake. This work is licensed under a Creative Commons Attribution 4.0 International License. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 2
3. Lecture notes —https://github.com/jj1bdx/ oueees-201806-public/ —Check out the README.md file and the issues! Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 3
4. Reporting —Keyword at the end of the talk —URL for submitting the report at the end of the talk Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 4
5. Today's topic: flexible packet routing and transport protocols Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 5
6. Various aspects of routing —Delivery —Addresses —Static or dynamic —Route aggregation —Security Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 6
7. Delivery schemes —Unicast —Broadcast/Multicast/Anycast Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 7
8. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 8
9. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 9
10. Internet protocol (IP) and IP addresses Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 10
11. Role of IP addresses —Network numbers —Interfaces: connected to the networks —Host IDs in the numbered networks —Global uniqueness —Special addresses (private, broadcast, multicast, loopback, etc.) Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 11
12. IPv4 addresses: 32 bits 192.168.100.20 In hexadecimal notation: 0xC0A86414 —4 x 0~255 numbers split with dots —Relatively easy to remember, but already being used up Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 12
13. IPv4 address with netmask 192.168.100.20/24 —Network: 192.168.100.0/24 —Host: number 20 (0~255) (32-24=8) —Host 0 = network itself —Host 255 = broadcast Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 13
14. Address in another netmask 192.168.100.20/28 —Network: 192.168.100.16/28 —Host: number 2 (0~15) (32-24=4) —Host 0 = network itself —Host 15 = broadcast —Different netmask = different address interpretation Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 14
15. Private addresses (RFC1918) No global routing for these address blocks —10.0.0.0/8 —172.16.0.0/12 (172.{16~31}.*.*) —192.168.0.0/16 (192.168.*.*) Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 15
16. Other special addresses (RFC6890) —0.0.0.0/8: "This" network —100.64.0.0/10: Shared address —127.0.0.0/8: Loopback —169.254.0.0/16: Link local —192.0.0.0/24: IANA specific —192.0.2.0/24, 198.51.100.0/24, 203.0.113.0/24: Documentation —192.88.99.0/24: 6to4 Relay Anycast —198.18.0.0/15: Benchmarking —240.0.0.0/4: Reserved —255.255.255.255/32: Limited broadcast Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 16
17. IPv6 addresses: 128 bits 2404:6800:400a:80c::2004 = 2404:6800:400a:080c:0000:0000:0000:2004 —www.google.com as of 2-JUL-2018 0905UTC —:xxxx: = up to 4 hex digits —:: = arbitrary number of 0, appearing only once in an address Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 17
18. IPv6 addresses with netmask 2404:6800:400a:80c::2004/64 —Network: 2404:6800:400a:80c::/64 —Host number: 0x0000000000002004 —Host number: 64 bits (0: network) —Broadcast -> multicast addresses —ff02::1 = all hosts, ff02::2 = all routers, etc. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 18
19. Why IPv4 to IPv6? —Because we've used up the 32-bit IPv4 addresses already —No more new address block for IPv4 —You need to buy unused blocks from other users —Took ~20 years (1996-2016) for the transition from IPv4 to IPv6 Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 19
20. Routing Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 20
21. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 21
22. Static routing —Set the default route for nodes which are not directly reachable —Works well on simple networks or star networks —Static routing may cause ping-pong Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 22
23. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 23
24. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 24
25. Dynamic routing —Hop count: count the hops between nodes —Link cost: determined by the speed and quality —Administrative policies Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 25
26. Simple hop counting —Assume every link costs the same with each other Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 26
27. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 27
28. Evaluating link cost —What if the cost of each link varies? —If two or more paths have the equal cost, all of the links will be utilized for load balancing Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 28
29. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 29
30. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 30
31. Simulating link failures —What if the link suddenly degrades or is disconnected? —Largely increasing the cost of degraded or disconnected links will give an easy solution Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 31
32. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 32
33. Administrative policies —For many reasons, you don't want to accept packets from some nodes, depending on the relay paths —For example: passing C is OK, but passing E is not: A-B-C-G and A-B-C-F- G are OK, but A-D-E-F-G is blocked —Common among interconnection of the autonomous systems (internet serpice providers and organizations) Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 33
34. Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 34
35. Routing information dissemination protocols —Link-state protocol: flooding link cost information of each node throughout the network —Path vector protocol: exchanging path of nodes for each network instead of the link costs —Highly vulnerable to external attacks Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 35
36. Routing aggregation —The following four networks —192.168.100.0/24 —192.168.101.0/24 —192.168.102.0/24 —192.168.103.0/24 —-> aggregated as 192.168.100.0/22 —4 networks together as one aggregated network Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 36
37. Network transports Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 37
38. IP address and the port number —Each service has a 16-bit port number —HTTPS = 443, DNS = 53, SSH = 22, etc. —A pair of IP address and port number defines an endpoint of communication Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 38
39. UDP and TCP —Two major transport protocols on the internet —User Datagram Protocol (UDP): connection-less —Transport Control Protocol (TCP): connection-oriented Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 39
40. Packet exchange limitation —Packets are not always delivered —Sending sequence is not preserved —The same packet may be received multiple times —The content of the packet may get altered or damaged —Packet size has the limitation Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 40
41. What UDP does —Add a header with the port number —Send it in an IP packet —... and that's it Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 41
42. UDP's pros and cons —UDP datagrams are still not always delivered and may get lost —Sequence is not preserved —The same datagram may be received multiple times and may cause duplicate delivery —The errors in the contents of UDP datagrams are detectable —UDP datagram has the size limit: suitable for relatively small messages —Very small additional latency Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 42
43. Transport control protocol (TCP) —Detect packet loss by timeout —Split stream into segments —Put sequence numbers to the segments —Reassemble segments to the stream —Perform congestion control Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 43
44. TCP's pros and cons —Loss is detected and recovered so long as the connection is alive —Sequence is preserved —No content repetition —Errors are detected and fixed by retransmission —The stream will accept data so long as the connection is alive —Data delivery may delay if retransmission occurs Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 44
45. Web: HTTP/2 (TCP) .vs. QUIC (UDP) —People wants speed —HTTP/2: stream aggregation and content compression —HTTP/2 is still bound by TCP —QUIC: tightly integrated to HTTP/2 and specific congestion control —Google is migrating to HTTP/2 + QUIC Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 45
46. Topics on next talk —Cloud computing .vs. endpoint computing —Sharing .vs. message passing —Centralization .vs. decentralization Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 46
47. Photo credits —All photos are modified and edited by Kenji Rikitake —Title: Denis Nevozhai via Unsplash —Unicast/broadcast/multicast/anycast diagrams: By Easyas12c~commonswiki / Perhelion, via Wikimedia Commons, CC0 (Public Domain) Kenji Rikitake / oueees 201806 part 2 3-JUL-2018 47