GO BIG, GO FAST, GO HOMEWhen Mesh Network Design MattersHarrison Chin, P.EngSenior Field Applications Engineer

OVERVIEW

0102

0304

0506

Introduction Test Methodology Reasoning

Relevant Use Cases Test Results Conclusion

© 2

019

Gar

min

Can

ada

Inc.

WHAT IS A MESH NETWORK?A system of interconnected nodes which cooperatively transport data from one node to another.

© 2

019

Gar

min

Can

ada

Inc.

WHY USE A MESH NETWORK?

Maximize Area Coverage

Nodes extend the effective range

Maximize Reliability

Redundant travel paths for data

Minimize Security Risks

Operate as independent networks

2

1 3 © 2

019

Gar

min

Can

ada

Inc.

USER INTERACTIVE SYSTEMSLatency sensitiveUse cases• Home automation• Lighting control• Door bells• Digital assistants

© 2

019

Gar

min

Can

ada

Inc.

USE CASELIGHTING CONTROL• Users expect instantaneous

response• Within 100ms or less

• Users expect consistency• Groups of lights change

together

© 2

019

Gar

min

Can

ada

Inc.

LIGHTING CONTROL SPECIFICATIONS• Bytes of data per packet• State data• On/Off• Intensity

• Infrequent events• Minutes between

changes• Strict latency bounds

© 2

019

Gar

min

Can

ada

Inc.

OBSERVATIONAL SYSTEMSThroughput sensitiveUse cases• Process management• Office space

management• Asset tracking• Vertical farm

monitoring©

201

9 G

arm

in C

anad

a In

c.

USE CASEASSET TRACKING• Hundreds of physical assets• Data is required in a regular and

reliable basis• Location, temperature, motion• Timescale in seconds to minutes

• Operate under a variety of environmental conditions

• Straightforward to deploy in scale

© 2

019

Gar

min

Can

ada

Inc.

ASSET TRACKING SPECIFICATIONS• Dozens of bytes per packet• Location• Sensor data

• Periodic events• Seconds between asset updates

• Large volume of concurrent traffic• All node data is important

© 2

019

Gar

min

Can

ada

Inc.

Use Case Latency Throughput Reliability

Lighting Control ✓ ✓

Asset Tracking ✓ ✓

KEY PERFORMANCE INDICATORS

© 2

019

Gar

min

Can

ada

Inc.

• Average Round Trip Time - Latency• Request a response from a node and measure how long it took

• Aggregate Gateway Throughput - Throughput• !"#$%& '( $)*%+ &%,%-.%/ (&'# 011 !'/%+

*'*01 *-#% ('& 011 +",,%++("1 &%+2'!+%+

• Average Success Rate - Reliability• !"#$%& '( &%+2'!+%+ &%,%-.%/*'*01 !"#$%& '( &%3"%+*+

© 2

019

Gar

min

Can

ada

Inc.

HOW DO WE MEASURE PERFORMANCE?

PING TESTING• Gateway-based Architecture• Phone/switch is the gateway for lighting• Cellular/Satellite/Hardline bridge for Asset tracking ©

201

9 G

arm

in C

anad

a In

c.

PING TESTING

© 2

019

Gar

min

Can

ada

Inc.

TEST FRAMEWORK

Bluetooth Mesh ANT BLAZE

nRF52832-based modules nRF52832-based modules

Certified nRF BLE Mesh 3.1.0 ANT BLAZE Libraries

Test software provided by Nordic Semiconductor Test software provided by Garmin

© 2

019

Gar

min

Can

ada

Inc.

TRAFFIC PATTERNSSingle Ping• One node responds• Lights…visible feedbackGroup Ping• 10 nodes respond• Many sensors relaying data to gateway

© 2

019

Gar

min

Can

ada

Inc.

LAYOUT - DENSE - 30 NODES

RelayNodeGateway

© 2

019

Gar

min

Can

ada

Inc.

DENSE – 30 NODESTest Setup• ~4m spacing between nodes• ~8m spacing between relays• 4 relays in 30 nodes

• Kitchen Environment• Reduced BLE ADV Traffic

© 2

019

Gar

min

Can

ada

Inc.

132.53

25.57

5.49 8.6619.97

33.33

138.66

192.39

0

50

100

150

200

250

5-Byte Single Ping 25-Byte Single Ping 5-Byte Group Ping 25-Byte Group Ping

Throughput (B/s)

Aggregate Throughput at Gateway (B/s)

Bluetooth Mesh ANT BLAZE

DENSE – 30 NODES

38

978

2592

7162

250

750

236

994

0

1000

2000

3000

4000

5000

6000

7000

8000

5-Byte Single Ping 25-Byte Single Ping 5 Byte Group Ping 25-Byte Group Ping

Round Trip Time (ms)

Average Round Trip Time (ms)

Bluetooth Mesh ANT BLAZE

DENSE – 30 NODES

© 2

019

Gar

min

Can

ada

Inc.

77.66

99.03

69.41

53.43

100 100 100 100

0

20

40

60

80

100

120

5-Byte Single Ping 25-Byte Single Ping 5-Byte Group Ping 25-Byte Group Ping

Success Rate (%)

Average Ping Success Rate (%)

Bluetooth Mesh ANT BLAZE

DENSE – 30 NODES

© 2

019

Gar

min

Can

ada

Inc.

LAYOUT – SPARSE – 30 NODES

© 2

019

Gar

min

Can

ada

Inc.

SPARSE – 30 NODESTest Setup• ~10m Spacing• 30 Nodes Total• All Relays• 100 Test Cycles• Increased BLE ADV Traffic

Logan and Rainer*May not be a representative

depiction of all interns

© 2

019

Gar

min

Can

ada

Inc.

76.3

32.28

13.764 10.6714

4147

120

0

20

40

60

80

100

120

140

5-Byte Single Ping 25-Byte Single Ping 5-Byte Group Ping 25-Byte Group Ping

Throughput (B/s)

Aggregate Response Throughput at Gateway (B/s)

Bluetooth Mesh ANT BLAZE

SPARSE – 30 NODES

© 2

019

Gar

min

Can

ada

Inc.

65.7

930.1

1241

3624

357610

1298 1360

0

500

1000

1500

2000

2500

3000

3500

4000

5-Byte Single Ping 25-Byte Single Ping 5-Byte Group Ping 25-Byte Group Ping

Round Trip Time (ms)

Average Round Trip Time (ms)

Bluetooth Mesh ANT BLAZE

SPARSE – 30 NODES

86.033

97.4333

70.17

29.17

100 100 100 99.6

0

20

40

60

80

100

120

5-Byte Single Ping 25-Byte Single Ping 5-Byte Group Ping 25-Byte Group Ping

Success Rate (%)

Average Ping Success Rate (%)

Bluetooth Mesh ANT BLAZE

SPARSE – 30 NODES

© 2

019

Gar

min

Can

ada

Inc.

Latency

Throughput Reliability

You can only pick two.

TRADE-OFF TRIAD!= CONJOINED TRIANGLES OF SUCCESS

© 2

019

Gar

min

Can

ada

Inc.

WHY? SHARING A LIMITED RESOURCE… AIR TIME• Fundamental contention between…• Keeping nodes busy with required data transfers• “Throughput”

• Keeping nodes available to listen for new data requests• “Latency”

• Keeping nodes aware of failed data transfers• “Reliability”

© 2

019

Gar

min

Can

ada

Inc.

Parameter Bluetooth Mesh ANT BLAZE

RF channel N/A, ADV channels 78 channels, Up to 3

App level retries SAR and/or model calls Library calls in app

Proxy ADV Interval Modify for density N/A, As needed

Transmission interval Modify for concurrency N/A, Built-in

Transport layer repeats Modify for interference N/A, Built-in

Relay count Modify for coverage N/A, All relay

Time-to-live Modify for coverage N/A

Bearer layer repeats Modify for interference N/A

TUNING PARAMETERS FOR THE RIGHT BALANCE

© 2

019

Gar

min

Can

ada

Inc.

LIGHTING CONTROL RE-IMAGINED

© 2

019

Gar

min

Can

ada

Inc.

BLUETOOTH MESH TIME DOMAINLIGHTING CONTROL

Gateway

Node 1

Node 2

Node 3

Node …

Time

© 2018 Garmin Canada

Newly published data0-20ms before

broadcast

Minimal activity between new data events

Node N

© 2

019

Gar

min

Can

ada

Inc.

ASSET TRACKING RE-IMAGINED

© 2

019

Gar

min

Can

ada

Inc.

BLUETOOTH MESH TIME DOMAINASSET TRACKING

Node 2

Node 3

Node …

Node N

Time

© 2018 Garmin Canada

Gateway

Node 1

Concurrent message collisions

© 2

019

Gar

min

Can

ada

Inc.

BLAZE BROADCAST TIME DOMAIN BOTH USE CASES

Gateway

Node 1

Node 2

Node 3

Node …

Steady state of activity between new data events

Regular interval between transmissions

62.5-500ms

© 2018 Garmin Canada

Newly sent dataWaits for next period

Node N

© 2

019

Gar

min

Can

ada

Inc.

“CONTEXT IS FOR KINGS”• What are your requirements?• Latency• Throughput• Device count

• Where are the nodes being deployed?• Commercial• Residential• Industrial

© 2

019

Gar

min

Can

ada

Inc.

- Hint: “Starfleet” Captain

CONCLUSION

• Is ultra-low latency required?• Data suggests look into Bluetooth Mesh

• Is increased throughput more important?• Data suggests look into ANT BLAZE

© 2

019

Gar

min

Can

ada

Inc.

QUESTIONS?

© 2

019

Gar

min

Can

ada

Inc.