NoCs 2 0 0 8 Zheng Shi and Alan Burns Real-time system group - - PowerPoint PPT Presentation

nocs 2 0 0 8
SMART_READER_LITE
LIVE PREVIEW

NoCs 2 0 0 8 Zheng Shi and Alan Burns Real-time system group - - PowerPoint PPT Presentation

Real-Time Communication Analysis for NoCs with Wormhole Switching NoCs 2 0 0 8 Zheng Shi and Alan Burns Real-time system group Department of computer science The University of York Outline 1 QoS in NoC 2 Priority based w orm hole sw


slide-1
SLIDE 1

Real-Time Communication Analysis for NoCs with Wormhole Switching

NoCs 2 0 0 8

Zheng Shi and Alan Burns

Real-time system group Department of computer science The University of York

slide-2
SLIDE 2

LOGO

Your site here

Outline

1 2 3

QoS in NoC

Priority based w orm hole sw itching W orst case netw ork latency analysis

4

Conclusion

slide-3
SLIDE 3

LOGO

Your site here

Netw orks on Chip ( NoC)

On-chip Communication:

Point-to-Point Bus

NoC: packet-switched, shared, optimized for communications

Resource efficiency High scalability IP reusability High performance

slide-4
SLIDE 4

LOGO

Your site here

NoC needs QoS

  • Differentiated Service Requirement
  • Best Effort
  • Guaranteed Service
  • Performance parameters:

latency, bandwidth, bounded jitter and loss probability, in-

  • rder data, etc.
  • Real-Time Service:
  • The correctness relies on not only the communication

result but also the completion time bound (deadline).

  • For hard real-time service, it is necessary that all the

packets must be delivered before their deadlines even under worst case scenario.

slide-5
SLIDE 5

LOGO

Your site here

Several Solutions

Contradiction:

The network gives more efficiency and flexibility but introduces the unpredictable delay due to the

  • contention. Real-time service, requires the timing to be

predictable even under the worst case situation

Contention avoidable

Circuit Switching : aSoC TDM : AEtheral, Nostrum

Contention acceptable

Priority based Wormhole Switching

slide-6
SLIDE 6

LOGO

Your site here

W orm hole Sw itching

Advantages (with Virtual Channels)

Small Buffer Size High Throughput Low Average Latency

slide-7
SLIDE 7

LOGO

Your site here

Priority Router Structure

There are sufficient VCs at each router Each VC is assigned distinct global priority Each flow also has distinct priority Flow only requests the VC with same priority At any time, only the flit with highest priority can access

the output link

Flit-level priority preemption between different VCs

slide-8
SLIDE 8

LOGO

Your site here

System Model

Characterize traffic-flow

  • A traffic-flow is packet stream which traverses the same

route from source to destination and requires the same grade

  • f service.

Attribute

P : Priority C : Basic network latency T : Period for periodic flow or minimal interval for sporadic flow D : Deadline

  • : Release Jitter

Interrelationship

Direct competing:

direct interference set:

Indirect competing:

indirect interference set

φ τ τ ≠ ∩ ) ( ) (

j i

Path Path

φ τ τ φ τ τ φ τ τ = ∩ ≠ ∩ ≠ ∩ ) ( ) ( , ) ( ) ( , ) ( ) (

k i k j j i

Path Path Path Path Path Path

} , ) ( ) ( , ) ( ) ( , ) ( ) ( | {

i j k k i k j j i k I i

P P P Path Path Path Path Path Path S > > = ∩ ≠ ∩ ≠ ∩ ∀ = φ τ τ φ τ τ φ τ τ τ

R

J

} , ) ( ) ( | {

i j j i j D i

P P Path Path S > ≠ ∩ ∀ = φ τ τ τ

slide-9
SLIDE 9

LOGO

Your site here

W orm hole Sw itching- A Case

1

τ

3

τ

2

τ

φ φ = =

I D

S S

1 1

,

φ τ = =

I D

S S

2 1 2

}, {

Priority ordering:

3 2 1

P P P > >

} { }, {

1 3 2 3

τ τ = =

I D

S S

slide-10
SLIDE 10

LOGO

Your site here

Characterize Netw ork Latency

Worst case network latency

:

The maximum length of time the packet could

take to travel from source to destination

The flow is schedulable if

Basic network latency C :

the network latency happens when there no traffic-flow contention exists.

S Hop B f f L L C

link size size add

⋅ + ⋅ ⎥ ⎥ ⎤ ⎢ ⎢ ⎡ + = /

max

R

D R ≤

slide-11
SLIDE 11

LOGO

Your site here

Model and Assum ption

The physical communication links are treated as shared competition resource At any time, only one traffic-flow is permitted to access the shared path The packet moves ahead when gets highest priority along the path The arrivals of higher priority flows are considered as preemption interference The allowable service time for a flow is all the time interval at which no higher priority flow competes for the same physical link

slide-12
SLIDE 12

LOGO

Your site here

Netw ork Latency Evaluation( 1 )

Worst Case Network Latency:

: worst case latency : maximum interference

the packets is supposed with maximum length and released at maximum rate

i i i

I C R + =

i

I

i

R

∈ ∀

⎥ ⎥ ⎥ ⎤ ⎢ ⎢ ⎢ ⎡ + =

D i

S j j j R j i i

C T J R I

slide-13
SLIDE 13

LOGO

Your site here

Netw ork Latency Evaluation( 2 )

  • Worst case network latency equation

The eqaution is solved using iterative technique Iterative starts with and terminates when

  • r

which denotes the deadline miss for this flow.

∈ ∀

⎥ ⎥ ⎥ ⎤ ⎢ ⎢ ⎢ ⎡ + + =

D i

S j j j R i i i i

C T J R C R

i i

C R =

n i n i

R R =

+1

∈ ∀ +

⎥ ⎥ ⎥ ⎤ ⎢ ⎢ ⎢ ⎡ + + =

D i

S j j j R i n i i n i

C T J R C R

1 i n i

D R >

+1

slide-14
SLIDE 14

LOGO

Your site here

Consider I ndirect I nterference ( 1 )

  • Minimal interval between subsequent

preemption is less than period

  • This could happen only when indirect interference is

considered. 3

τ

1

τ

i

T

i

R

2

τ

slide-15
SLIDE 15

LOGO

Your site here

Consider I ndirect I nterference ( 2 )

  • Preemption interference upper bound

Worst case latency

∈ ∀

⎥ ⎥ ⎥ ⎤ ⎢ ⎢ ⎢ ⎡ − + + =

D i

S j j j j j R j i i

C T C R J R I

j

τ

i

τ

j

R

R j

J

i

R

j

T

j

C

I j

J

j

T

j

T

j

T

∈ ∀

⎥ ⎥ ⎥ ⎤ ⎢ ⎢ ⎢ ⎡ − + + + =

D i

S j j j j j R i i i i

C T C R J R C R

slide-16
SLIDE 16

LOGO

Your site here

Case Exam ple

  • For : there is no higher priority flow

than , so

  • For : shares the physical link with

higher priority flow and

Trafffic- Flow s

C P T D 2 1 6 6 3 2 7 7 3 3 13 13

2

τ

3

τ

1

τ

1

τ

2

1 1

= = C R

2

τ

1

τ

2

τ

1

τ

φ τ = =

I D

S S

2 1 2

}, {

  • suffers both direct and indirect

interference with

The interference jitter of referred to equals So which stops at 5 2 6 5 3 5 2 6 3 3 3

2 2 1 2 2

= ⎥ ⎥ ⎤ ⎢ ⎢ ⎡ + = = ⎥ ⎥ ⎤ ⎢ ⎢ ⎡ + = = R R R

3

τ

} { }, {

1 3 2 3

τ τ = =

I D

S S

2

τ

3

τ

2 3 5

2 2

= − = −C R

2 2 2 2 3 3 3

C T C R R C R ⎥ ⎥ ⎤ ⎢ ⎢ ⎡ − + + = 9

3 =

R

slide-17
SLIDE 17

LOGO

Your site here

Tightness of analysis ( 1 )

1

τ

2

τ

3

τ

φ φ = =

I D

S S

1 1

,

φ φ = =

I D

S S

2 2

,

φ τ τ = =

I D

S S

3 2 1 3

}, , {

3 2 1

P P P > >

Priority ordering:

slide-18
SLIDE 18

LOGO

Your site here

Tightness of analysis ( 2 )

  • Parallel Interference
  • When parallel interference exists, the real worst case network

latency is no more than the analysis result.

  • When parallel interference exists, finding worst case network

latency is NP-hard (the proof details refers the paper).

  • Our analysis is safe but pessimistic.

3

τ

1

τ

2

τ

slide-19
SLIDE 19

LOGO

Your site here

Conclusion

Real time communication service can be supported by priority based wormhole switching technique. The schedulable test is derived by worst case network latency analysis. Both direct and indirect interferences are taken into account. When parallel interference exists, finding worst case network latency is NP-hard, but our analysis still form an upper bound.

slide-20
SLIDE 20

Thank you… and Question