32
Balaji Prabhakar Spring 2011 A history of big routers (slides from Nick McKeown’s EE 384X presentation) EE384M Network Algorithms

Balaji Prabhakar

  • Upload
    tayten

  • View
    44

  • Download
    4

Embed Size (px)

DESCRIPTION

Spring 2011 A history of big routers (slides from Nick McKeown’s EE 384X presentation). EE384M Network Algorithms. Balaji Prabhakar. Outline. What is an Internet router? What limits performance: Memory access time The early days: Modified computers Programmable against uncertainty - PowerPoint PPT Presentation

Citation preview

Page 1: Balaji Prabhakar

Balaji Prabhakar

Spring 2011A history of big routers

(slides from Nick McKeown’s EE 384X presentation)

EE384MNetwork Algorithms

Page 2: Balaji Prabhakar

OutlineWhat is an Internet router?

– What limits performance: Memory access time

The early days: Modified computers– Programmable against uncertainty

The middle years: Specialized for performance– Needed new architectures, theory, and

practice– So how did we do?– Simple model breaking down

Page 3: Balaji Prabhakar

Definitions

12

34

56

78

……

N

R

N = number of linecards. Typically 8-32 per chassisR = line-rate. 1Gb/s, 2.5Gb/s, 10Gb/s, 40Gb/s, 100Gb/s

Capacity of router = N x R

Page 4: Balaji Prabhakar

What a Big Router Looks LikeCisco GSR 12816 Juniper T640

6ft

19”

2ft

Capacity: 640Gb/sPower: 5kW

3ft

2.5ft

19”

Capacity: 320Gb/sPower: 3kW

Page 5: Balaji Prabhakar

What Multirack Routers Looks Like

Cisco CRS-1 Juniper T1600 + TX Matrix

Page 6: Balaji Prabhakar

Lookup internet addressCheck and update age

Check and update checksum

Page 7: Balaji Prabhakar

Router Control and Management

Barebones Router

Page 8: Balaji Prabhakar

Barebones Router

Page 9: Balaji Prabhakar

Barebones Router

Page 10: Balaji Prabhakar

1 2

BottlenecksMemory, memory, …

Page 11: Balaji Prabhakar

OutlineWhat is an Internet router?

– What limits performance: Memory access time

The early days: Modified computers– Programmable against uncertainty

The middle years: Specialized for performance– Needed new architectures, theory, and

practice– So how did we do?– Simple model breaking down

Page 12: Balaji Prabhakar

Early days: Modified Computer

R

R

R

R

R

R

R

R

Must run at rate N x R

Bottlenecks

Page 13: Balaji Prabhakar

2nd Generation RouterR

R

R

R

Page 14: Balaji Prabhakar

Early days: Modified Computer

Function more important than speed

1993 (WWW) changed everything

We badly needed– Some new architecture– Some theory– Some higher performance

Page 15: Balaji Prabhakar

N x R

3rd Generation Router: Switch

Page 16: Balaji Prabhakar
Page 17: Balaji Prabhakar

1 x R

Arbiter

Page 18: Balaji Prabhakar

Arbiter Arbiter Arbiter Arbiter

Arbiter Arbiter Arbiter Arbiter

Arbiter

Page 19: Balaji Prabhakar

4th Generation RouterMultirack; optics inside

SwitchLinecards

Optical links

100sof metres

Page 20: Balaji Prabhakar

Alcatel 7670 RSP Juniper TX

TX

More 4th Generation Routers

Avici TSR Cisco CRS-1

Page 21: Balaji Prabhakar

Example of Theory

There’s something special about “2”

Page 22: Balaji Prabhakar

Case 1: Placing calls

A crosspoint switch supports all permutationsSo it is “non-blocking”But it needs N2 crosspoints

1

Permutation

1 0 0 0 0 0 0 00 0 0 0 0 0 00 0 0 0 1 0 0 00 0 0 1 0 0 0 00 0 1 0 0 0 0 00 0 0 0 0 1 0 00 0 0 0 0 0 0 10 0 0 0 0 0 1 0

1

11

1 11

1

11

Crosspoint switch1

11

11

1

1

Page 23: Balaji Prabhakar

Case 1: Placing CallsUncertainty costs

1 0 0 0 0 0 0 00 1 0 0 0 0 0 00 0 0 0 1 0 0 00 0 0 1 0 0 0 00 0 1 0 0 0 0 00 0 0 0 0 1 0 00 0 0 0 0 0 0 10 0 0 0 0 0 1 0

00000000

0 0 0 0 0 0 0 0 1

If I give you the permutation, you can route it. If I give you entries one at a time, you can’t.

1

1 1

1

Clos (1950s): But if you make it run 2 times faster

you can route calls one at a time.

Page 24: Balaji Prabhakar

Case 2: Mimicking

N x R

Page 25: Balaji Prabhakar

Case 2: Mimicking

1 x R

Page 26: Balaji Prabhakar

Are they equivalent?

NR

R

No.

Page 27: Balaji Prabhakar

Case 2: Mimicking

1 x R ? x R

Algorithm

Page 28: Balaji Prabhakar

Yes, if it runs 2 times faster.

Now are they equivalent?

R 2R

Algorithm

NR

Page 29: Balaji Prabhakar

Yes, if it runs 2 times faster.

Case 3: Are they equivalent?

Page 30: Balaji Prabhakar

1

Case 4: Routing packets with uncertainty

R

Algorithm

0.1 0.2 0.5 0.20.3 0.1 0.3 0.30.5 0.2 0.1 0.20.1 0.5 0.1 0.3

Rates

But we don’t know the rates (they are always changing)

If you know the rates, you can find a sequence of permutations:

0 0 1 00 1 0 00 0 0 11 0 0 0

0 0 1 01 0 0 00 0 0 10 1 0 0

1 0 0 00 0 1 00 0 0 10 1 0 0

=

Page 31: Balaji Prabhakar

Case 4: Routing packets with uncertainty

2If you choose the permutations one at a time, and you can spend as long as you want choosing, then you can support any pattern of rates.

3But if you have to make decisions one at a time, then the switch has to run 2 times faster.

Page 32: Balaji Prabhakar

Case 5: Load-balancing Load-balancing to support all rate matrices:– Requires the network to run 2 times faster– E.g. the VL2 (Valiant Load balancing)

architecture for Data Centers