Scribd
Upload
18 views38 pages

Lect15 Internetarch

Uploaded by

olajameel89
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
18 views38 pages

Lect15 Internetarch

Uploaded by

olajameel89
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 38

Internet Architecture

ECE/CS598HPN
Radhika Mittal
What is Internet Architecture?

• How an endhost identifies and specifies the destination.


• How routers understand that specification to forward
packets to the destination over the Internet.

• Carried out by L3 (IP).


IP as the narrow waist

• Facilitated a lot of innovation


above and below IP.

• Hard to change IP itself.


Security

• Clean-slate architecture.

• Establishes trust domains.

• Guarantee control plane


isolation for trust domains.
Security
• Architecture to limit DoS
attacks.

• Receivers grant sending


capabilities to senders.

• Routers check for this


capability to determine if
the packet is wanted.
Accountability

• Make Internet addressing


more accountable.

• Use self-certifying host


addresses.
Information-centric Networking

• Name bits (data or


content) instead of
locations.

• Self-certified or signed in
some manner.
Other forms of addressing

Replace IP addresses with Replace IP addresses (and


location-independent, flat, ports) with service names.
globally unique IDs.
Source routing
Lots of proposals
How do we enable innovation in
Internet Architecture?
Trotsky: Enabling a Permanent
Revolution in Internet Architecture
James McCauley, Yotam Harchol, Aurojit Panda,
Barath Raghavan, Scott Shenker

SIGCOMM’19

Some slide contents borrowed from McCauley’s SIGCOMM’19 talk.


How do we enable innovation in
Internet Architecture?

• Remove the narrow waist!

• How?

• Two steps
Step 1: Fix Layering

• We are missing a layer!

• Internet is not a composition


of L2 networks.

• It is a composition of domains.
Step 1: Fix Layering

• We are missing a layer!

• Internet is not a composition


of L2 networks.

• It is a composition of domains.
Step 1: Fix Layering

• Decouple how data is


delivered:
• within a domain (L3)
• across domains (L3.5)

• Decouple how two


domains deliver data
internally.
How do we enable innovation in
Internet Architecture?

• Remove the narrow waist!

• How?

• Two steps:
• Layer 3.5: decouple intra-domain and inter-domain data delivery.
Step 2: Embrace multiple architectures

• Support multiple L3.5 protocols.


• Up to the domain to choose which ones it wants to
support.

• Trotsky Processors (TPs) deployed at domain edge (in


software) responsible for implementing supported L3.5
protocols.
How do we enable innovation in
Internet Architecture?

• Remove the narrow waist!

• How?

• Two steps:
• Layer 3.5: decouple intra-domain and inter-domain data delivery.
• Embrace multiple L3.5 protocols instead of upgrading to a single
one.
Inside a domain
Inside a domain
Inside a domain
Inside a domain
Host initialization
Host initialization
Host initialization
Host initialization
Host initialization
Host initialization
Web Download
Web Download
Partial Deployment of L3.5 Designs

Domain A --- Domain B ---- Domain C


L3.5 L3.5 L3.5 L3.5
Key Contribution of Trotsky

• A framework that allows incremental side-by-side


deployment of new architectures.

• And is itself incrementally deployable.


Summary
• Goal: enable extensibility in Internet architecture.

• Problem: the universal narrow waist.

• Solution: remove it!


• Decouple intra-domain and inter-domain data planes.
• Embrace co-existence of multiple inter-domain protocols.

• Result:
• An incrementally deployable design.
• ..which can incrementally deploy new architectures.
Discussion

• Is the universal narrow waist truly removed?

• What are the limitations of Trotsky design?


Your opinions

• Pros
• Backwards-compatible and incrementally deployable.
• Framework providing only a minimal set of functionality.
• No need to change all routers.
• Opens up avenue for future research.
Your opinions
• Cons
• Overhead of mapping L3.5 to/from underlying layers.
• Overhead of implementing an L3.5 protocol (in software).
• Pairwise translators needed at domain edge.
• Requires some form of cooperation between ASes.
• Can a network middlebox provide the same functionality at
Trotsky?
• How crucial is the decoupling between L3 and L3.5?
• To what extent can it provide security?
• Initial deployment is challenging.
• “Simplicity is a feature not a bug” – do we really need more
complex Internet architectures?
Your opinions

• Ideas
• Is Trotsky against end-to-end argument?
• Design DDoS resilient network architecture.
• Implementation and evaluation of L3.5 protocols.
• Why not implement Trotsky Processors in programmable switches?
• What are the limitations of proposed L3 protocols?
• Explore what incentivizes domains to support an L3.5 protocol.
• Experiment testbed that allows multiple architecture to co-exist.

You might also like