Tesseract: A 4D Network Control Plane

1
Tesseract A 4D Network Control Plane

• Sam McIngvale
• Yu Zhang

2
Why 4D?

• Why not 3D or 5D?

3
Single Point-of-Failure

• Tesseract design essentially reduces the network
  to a single DE.
• If this DE fails or is compromised, the entire
  network is at stake.
• Even with multiple DEs, the network is at a
  greater risk because there always a single
  point-of-failure

4
Efficiency of the Decision Plane

• Maybe efficient on single link failure.
• Different story when multiple failures happen.
• How many events will be collected during the
  hold-down timers?

5
Overall Efficency p.1

• Currently, Tesseract is already slower than fast
  OSPF
• Lots of encryption
• All dessemination plane traffic is
  encrypted/decrypted at each node
• All BGP traffic is encrypted and must be sent to
  the DE
• All DHCP reqests are encrypted and must be sent
  to the DE

6
Overall Efficency p.2

• Tesseract not tested in the wild
• Convergence times could actually be much slower
  given the fact that end-host traffic was not
  present in simulations
• Network can bottleneck at DE
• Thus, the network can only be as fast as the DE
  allows it to be.

7
Emulation and Simulation

• Can it represent whats happening in the real
  world?
• The time required for the entire network to
  reconverge after an event.
• Until Tesseract is tested in the wild, we will
  not know how it performs under the strains of
  real-world traffic

8
Do not give better performance!
9
Overhead Increases Exponentially
10
Scalability

• Can it scale to todays Internet?
• Tomorrow?