technische universität münchen on the quantification of sustainability and extensibility of...

11
Technische Universität München Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit Chakraborty TU Munich, Germany DAC’2011, San Diego, US, June 5-10, 2011 Unmesh Bordoloi, Petru Eles, Zebo Peng Linkoeping University, Sweden

Upload: virgil-barnett

Post on 17-Jan-2018

216 views

Category:

Documents


0 download

DESCRIPTION

Technische Universität München 1 Latest minislot to transmit m 3 B 1 = 1 R 1 = 2 S 1 = 6 FlexRay Protocol cycles slots m1m1  Transmission points of m i are uniquely specified by the tuple S 1 = m3m3 m3m3 pLatestTx=6 displaced m2m2 m2m2  Communication is organized in a periodice sequence of bus cycles  Static (time-triggered) and dynamic (event-triggered) segments  Static slots denote time windows of fixed and equal length  Dynamic slots are logical entities which specify priorities and comprise of several minislots depending on the message sizes static dynamic Static slots minislots and m4m4 m4m4 m4m4 m4m4 Examples:

TRANSCRIPT

Page 1: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität MünchenTechnische Universität München

On the Quantification of Sustainabilityand Extensibility of FlexRay Schedules

Reinhard Schneider, Dip Goswami, Samarjit Chakraborty

TU Munich, Germany

DAC’2011, San Diego, US, June 5-10, 2011

Unmesh Bordoloi,Petru Eles, Zebo Peng

Linkoeping University, Sweden

Page 2: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

Automotive Design Process

2

Protocol configurationInitial design phase

FlexRay Network

Incremental scheduling

New applications at each design cycle

Scheduling

Update

Existing schedules

+

Physical layer configuration

Will deadlines of existing messages be met in presence of future messages?

Are enough suitable schedules available to accomodate future messages? How to quantify and interpretate sustainable and extensible schedules? How to capture all FlexRay-specific properties in an analysis framework?

?

?

Page 3: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

1

Latest minislot to transmit m3

B1 = 1

R1 = 2

S1 = 6

FlexRay Protocol

3

.

..

cycles

1 2 3 4 5 slots

0

2

4

62

.

..

m1

Transmission points of mi are uniquely specified by the tuple

S1 = 7

1 2 3 4 5 6 7 8

m3

m3

pLatestTx=6

displaced

m2

m2

Communication is organized in a periodice sequence of bus cycles

Static (time-triggered) and dynamic (event-triggered) segments

Static slots denote time windows of fixed and equal length

Dynamic slots are logical entities which specify priorities and comprise of several minislots depending on the message sizes

static dynamic

Static slots

minislots

63

3

5

and

m4

m4

m4

m4

Examples:

Page 4: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

a) Deadline d3 of m3 is still metin presence of m6? cycle

slot

0

1

2

3

4

5

62

63

.

..

m1

m1

m1

m1

m2

m2

m2

m3

m3

m3m4

m4

4

m6 Design cycle I+1m5

m2 m3m6

Sustainability Analysis

1 2 3 4 5

m7

m7

c) How to estimate the workloadof future messages?

b) Sufficient minislots availableto transmit m3?

m5

m5

m5

m5

Workload estimation

bus

Intuitive idea: e.g., f(n)=3

n1

busn2

busn3

n1=6 bytes

n2=8 bytes

n3=10 bytes

Several payload sizes n result in a constant minislot consumption f(n) on the bus

From the FlexRay specification B.4.14 [13]:

Temporal isolation between slots in the static segment

Interference of messages withhigher priorities in the dynamic segment

Design cycle I

Sustainability test:

Page 5: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München 5

1 2 3 4 5 6

cycle

0

1

2

3

4

5

62

63

.

..

m1

m1

m1

m1

m2

m2

m2

m2

m3

m3

m3

m3

b) Slots might be reserved for special functionalities

Extensibility Analysis

m4

m4

Quality rating function P1(Si): Ability of a slot Si to provide real-time guarantees (priorities)

Grade of extensibility P2(Si): Ability of a slot to provide versatile schedules to accomodate future messages

a) How to measure extensibility for a particular slot?

Empty slotOnly particular cycles available

c) How to quantify priorities of slots

Page 6: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

Extensibility Index

6

Holistic quantification of extensibility for FlexRay networks that depends on both, the Grade of Extensibility and the Quality Rating

Benefits: Only one metric necessary to quantify and compare the extensibility

of FlexRay networks

Easy visualization and interpretation

Applicable to static and dynamic segments

Page 7: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

Experimental Results

7

Visualization of the extensibility index Quality rating function 4 minislots considered for future messages, i.e., payload 22 – 38 bytes

d10 = 22msD10 = 11,30ms

In future: µ10 = 317 > 237

µ10 = 167pLatestTx = 237

d94 = 22msD94 = 21,56ms

µ94 = 152pLatestTx = 237

In future: D94 = 22,76 ms > 22ms

empty slots: E(Si)=P1(Si)

completely filled slots: E(Si)=0

Both messages, m10 and m94 meet their constraints now (at current design iteration) but

do not pass the sustainability test

Full available static slots

Unavailable static slots

Full available dynamic slots with high priorities

Unavailable dynamic slotswith high priorities

Only some scheduleswith high priorities available

Page 8: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

Concluding Remarks

8

Notion of sustainability and extensibility in the context of FlexRay

Presented analysis framework reflects all protocol details

Easy visualization and interpretation

Future work envisages to automatically synthesize sustainable and extensible schedules

More details available at the poster session

Page 9: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

Backup for Q&A

9

Page 10: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

Grade of Extensibility

10

.

..

1 2 3 4 5 6 7 8 9 Si

2

3

49,6% of all schedules are available for future messages in slot 324,4% of all schedules are available for future messages in slot 6No schedules are available for future messages in slot 7All schedules are available for future messages in slot 849,6% of all schedules are available for future messages in slot 9

There are 127 choices to schedule a message mi in slot Si

10

slot...

cycle

4

5

62

63

.

..

..

.

cycle

1

0

4

5

62

63

.

..

Page 11: Technische Universität München On the Quantification of Sustainability and Extensibility of FlexRay Schedules Reinhard Schneider, Dip Goswami, Samarjit

Technische Universität München

Quality Rating Function

11

k

Example:

Reserved slots are not considered in the extensibility analysis, P1(Si)=0

Static slots are considered with highest priority, P1(Si)=1

Dynamic slots are considered with decreasing priorities, P1(Si) 0 as Si 258