+ All Categories
Home > Documents > Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1...

Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1...

Date post: 24-Feb-2021
Category:
Upload: others
View: 5 times
Download: 0 times
Share this document with a friend
16
Carbonaceous chondrites as bioengineered comets 1 Robert B. Sheldon and 2 Richard Hoover 1 Grassmere Dynamics LLC, 320 Sparkman Dr, Huntsville, AL, USA and 2 NASA/MSFC and Athens State University, Athens, AL ABSTRACT The discovery of microfossils on carbonaceous meteorites has electrified the public with the first concrete evidence of extraterrestrial biology. But how these organisms colonized and grew on the parent body–the comet–remains a mystery. We report on several features of cyanobacteria that permit them to bioengineer comets, as well as a tantalizing look at interplanetary uses for magnetite framboids that are found in abundance on carbonaceous chondrites. We argue that these structures provide important directionality and energy harvesting features similar to magnetotactic bacteria found on Earth. 1. INTRODUCTION The presence of cyanobacterial fossils on carbonaceous chondrites–black, crumbly meteorites widely believed to be extinct comets–suggest that not only has bacterial life thrived in extraterrestrial environments, 1 but that their growth has modified their cometary environment. 2 In a previous papers 2–4 we document a number of modifications that cyanobacteria can make to their cometary home: they can provide a polysaccharide binder (slime) that increases the tensile strength of the chondritic or granular matrix; polysaccharides blacken in the presence of ultraviolet light to lower the surface albedo; polysaccharide layers rupture at the high temperatures of the subsolar point to locally generate steam jets that provide an ”anti-stellar-accretion” force; and polysaccharides lower the freezing point of pure water so as to provide a longer growing season. All these modifications of the environment are a result of the extracellular polysaccharide sheaths, which cyanobacteria produce in abundance, but whose presence must be inferred from the ubiquitous kerogenic carbon microfossils of the original biomolecules created millions of years ago. By way of contrast, magnetite grains are essentially unchanged since their origin, and if they are created biologically, 5, 6 preserve their unique structure for millions of years. 7 We argue that the magnetite grains in carbonaceous chondrites are even stronger evidence than keragen that not only were they created biologically, but that they modify the cometary environment in subtle ways to enhance the growing season and perhaps even spread life more efficiently to other galaxies. Therefore we first present evidence that the magnetite is biological, we then compare cometary bio-magnetite with terrestrial bio-magnetite, and finally we speculate on the unique properties of cometary magnetite. 1.1. Terrestrial 1.1.1. The presence of biologically produced magnetite inclusions The literature on this topic is very recent, since the discovery that biology forms magnetite was only made in 1962, 5 and the discovery of magnetotactic bacteria was only made in 1975. 6 Since then the field of bio- magnetism has exploded. 8–12 Not only do bacteria orient themselves with chains of spontaneously magnetized, stable single-domain magnetite (SSD), but eukaryotic algae orient with multiple magnetosome chains, 13 and bees, bats, whales, 14 fish (trout), and homing pigeons 15 are all thought to use the compass-needle effect of a long chain of SSD magnetite. Other magnetic transducers have been examined, 16 but without any evidence that biology uses these weaker responses. Further author information: E-mail: [email protected], Telephone: 1 256 653 8592
Transcript
Page 1: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

PUBLIC

NXP, THE NXP LOGO AND NXP SECURE CONNECTIONS FOR A SMARTER WORLD ARE TRADEMARKS OF NXP B.V.

ALL OTHER PRODUCT OR SERVICE NAMES ARE THE PROPERTY OF THEIR RES PECTIVE OWNERS. © 2020 NXP B.V.

I E E E E t h e r n e t & I P Te c h D a y s – S e p t e m b e r 2 0 2 0

Don Pannell, Fellow

Automotive Ethernet Networking, NXP Semiconductor

Nicolas Navet, Professor

University of Luxembourg / Cognifyer.ai

Practical Use Cases for

Ethernet Redundancy

Page 2: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1PUBLIC

OVERVIEW

• Examined Redundant Network Configurations

• Soft Error Rate Modeling

• Hard Error & Cost Factor Considerations

• Summary

Page 3: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

2PUBLIC

THE NEED

➢Zonal networks, as shown in the figure, easily support redundancy, especially on the highlighted Ethernet backbone

➢IEEE 802.1CB is the TSN standard for Seamless Redundancy, supporting zero recovery time from lost frames

➢This presentation models CB in various topologies with the CB function at various locations in the network so these differences can be quantified

➢This is NOT a Safety presentation, but the data presented will help Network & Safety teams develop cost effective redundancy solutions

c

3

5

3

C

C

2

C

C

CC

2C

C

C

2

C

3

2

5

C

C

C

2

C

C

C

C

C

cC 2

C

2

CC

C C

C C C C

2

2 C5 2

Zonal

Page 4: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

3PUBLIC3

PUBLIC

NXP, THE NXP LOGO AND NXP SECURE CONNECTIONS FOR A SMARTER WORLD ARE TRADEMARKS OF NXP B.V.

ALL OTHER PRODUCT OR SERVICE NAMES ARE THE PROPERTY OF THEIR RES PECTIVE OWNERS. © 2020 NXP B.V.

Examined Redundant

Network ConfigurationsSingle changes to a baseline are compared so their effect can be

extrapolated

Page 5: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

4PUBLIC

4 CB Switches, Non-CB Talker & Listener

• Pros:

− Existing end-nodes can be used

− Hard or Soft errors on the backbone wires are protected

▪ A Hard error is a long duration error like a broken wire

▪ A Soft error is an intermittent error like a CRC errored frame

▪ The Brown & Blue wires are redundant paths for the packets

− Failure of S-B or S-Y is protected

• Cons:

− Backbone bandwidth is double+ for the redundant flows

▪ The ‘+’ is due to the added 6-byte R-Tag & possible 4-byte S-Tag

− Links from the Talker & to the Listener are not protected

▪ The Black wires

− Failure of S-A, T-1, S-Z or L-1 is not protected

T-1

S-A

L-1

S-B

S-Y S-Z

= CB Seq # & Split = CB Merge

Page 6: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

5PUBLIC

6 CB Switches, Non-CB Talker & ListenerT-1

S-A

L-1

S-B

S-C

S-Y

S-D

S-Z

• Very similar Pros & Cons as the previous slide

• Only change is the addition of S-C & S-D in the backbone

• Modeled to see the impact of Soft errors on the increased

number of links in the protected backbone

= CB Seq # & Split = CB Merge

Page 7: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

6PUBLIC

Non-CB Switches, CB Dual-Homed Talker &

ListenerT-1

S-A

L-1

S-B

S-Y S-Z

• Pros:

− Existing switches can be used

− Hard & Soft errors on the entire path are protected

▪ The Brown & Blue wires (dotted wires are not used for these flows)

− Failure of any single switch is protected

− Failure of anything in the Blue path (wire or switch) is protected

− Failure of anything in the Brown path is protected

− Backbone bandwidth is half of Network #1 for the redundant

flows as dotted wires are not used & are available

• Cons:

− Requires dual-homed end nodes (with dual Ethernet ports)

− End nodes replicate frames & eliminate the duplicates

− Failure of T-1 or L-1 is not protected

= CB Seq # & Split = CB Merge

Page 8: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

7PUBLIC

Non-CB Switches, CB Single-Homed Talker &

ListenerT-1

S-A

L-1

S-B

S-Y S-Z

• Pros:

− Existing switches & end nodes w/new software can be used

− Hard or Soft errors on the backbone wires are protected

▪ The Brown & Blue wires between the switches

− Soft errors on the links from the Talker & to the Listener are

protected

▪ Temporally due to the doubled transmission

− Failure of S-B or S-Y is protected

• Cons:

− End-to-end bandwidth is double+ for the redundant flows

▪ The ‘+’ is due to the added 6-byte R-Tag & possible 4-byte S-Tag

− End nodes replicate frames & eliminate the duplicates

− Failure of S-A, T-1, S-Z or L-1 is not protected

= CB Seq # & Split = CB Merge

Page 9: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

8PUBLIC

Mixed Switches, Mixed Talker & ListenerT-1

S-A

L-1

S-B

S-Y S-Z

• Very similar Pros & Cons as previous slide

• Only change is S-Z does the duplicate frame removal

− This means the S-Z to L-1 link is no longer protected

− But existing ECU’s without any software changes can be used

▪ With increased redundancy support achieved via firmware updates

▪ i.e., Network #1 → Network #5 → Network #4

▪ CB enabled switches are needed in this case

• Modeled to see the impact of Soft errors on the

unprotected link to the Listener

• This mixture is supported as long as T-1 creates the

Brown & Blue frames in accordance to 802.1CB

= CB Seq # & Split = CB Merge

Page 10: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

9PUBLICNXP, THE NXP LOGO AND NXP SECURE CONNECTIONS FOR A SMARTER WORLD ARE TRADEMARKS OF NXP B.V.

ALL OTHER PRODUCT OR SERVICE NAMES ARE THE PROPERTY OF THEIR RES PECTIVE OWNERS. © 2020 NXP B.V. 9

PUBLIC

Soft Error Rate Modeling

End-to-end protection is key for high-integrity communication

Page 11: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 0PUBLIC

Error model used for soft errors

• Soft errors are limited to frames dropped due to CRC errors

• Bit Error Rate (BER) is assumed identical on all links & constant over time

• CRC errors are independent of each other, i.e., no “bursts” of errors

• 100BASE-T1 specifies BER ≤ 10-10, PHYs are much better in practice, thus

a BER of 10-12 is used in the experiments

With 100BASE-T1 BER = 10-12, the average

time between 2 CRC errors on a link is

≈18h13m at 20% load and ≈3h38m at 100% load with minimum frame size

Page 12: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 1PUBLIC

Replication solutions : requirements & Single Points of Failure

1 2 3 4 5

2 CB-aware

switches2 CB-aware

switches w/more hops

CB-aware

talker & listener

+ dual-homing

CB-aware

talker & listener

+ temporal redundancy

CB-aware talker

+ 1 temporal redundancy

+ 1 CB-aware switch

+

= unprotected transmissions on that link → Single Point of Failure for Soft Errors

[RTa

W-P

eg

ase

scre

en

sh

ots

]

T-1 T-1

L-1L-1 L-1

S-B S-B S-B S-B S-BS-AS-A

S-DS-C

S-YS-Y S-Y S-ZS-Z S-Y

S-A

S-Y

= CB Seq # & Split = CB Merge

Page 13: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 2PUBLIC

Replication in action - packet loss rate

• Packets are lost when none of the copies are received by the listener(s)

• The data assumes a homogeneous Bit Error Rate = 10-12 regardless of link speed

• 2.44E-8 means 2.44 × 10-8

Network Loss Rate for

1522B Packet

Loss Rate for

64B Packet

Improvement factor

vs. No RedundancyTakeaways

2.44E-8 1.02E-9 2The two unprotected transmissions

are by far the dominant risk factor

2.44E-8 1.02E-9 2.5# of hops is a low order factor

as long as transmissions are protected

1.33E-15 2.36E-18 8.7E8 Very robust to soft errors!

2.37E-15 4.19E-18 4.9E8 Same order of magnitude as #3

1.22E-8 5.12E-10 4Twice as robust compared to #1 due to

the single unprotected transmission link

1

2

3

4

5

3 (dual-homing) and 4 (end-to-end temporal redundancy) stands out

additionally protects against any one hard error unlike

3 4

3 4

Page 14: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 3PUBLIC

Replication in action – average time between 2 packet losses

• Assuming a transmission period of 1ms with min. size frames (e.g., actuator messages)

• The data assumes a homogeneous Bit Error Rate regardless of link speed

• Stated times are for one flow only with the low link utilization as stated above

Network BER = 10-11 BER = 10-12 BER = 10-13 BER = 10-14

Baseline: no

redundancy13hours 33min 5days 15hours 56days 12hours ≈ 1year 6months

1day 3hours 11days 7hours 113days ≈ 3years 1month

1day 3hours 11days 7hours 113days ≈ 3years 1month

1.3E5 years 1.3E7 years 1.3E9 years 1.3E11 years

7.5E4 years 7.5E6 years 7.5E8 years 7.5E10 years

2days 6hours 22days 14hours 226days ≈ 6years 2months

1

2

3

4

5

Page 15: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 4PUBLIC1 4

PUBLIC

NXP, THE NXP LOGO AND NXP SECURE CONNECTIONS FOR A SMARTER WORLD ARE TRADEMARKS OF NXP B.V.

ALL OTHER PRODUCT OR SERVICE NAMES ARE THE PROPERTY OF THEIR RES PECTIVE OWNERS. © 2020 NXP B.V.

Hard Error & Cost

Factor Considerations

Nothing is free, and not all solutions cost the same!

Page 16: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 5PUBLIC

Cost Factor – Components

• Creating a Redundant backbone:

− 1st: Create a Ring network from a daisy-chain one

▪ Cost = 1 extra link only in the network

− 2nd: Add seamless redundancy: 802.1CB support in bridges in the critical data’s path

▪ Doubles the backbone’s bandwidth for the critical flows

▪ Cost = varies depending on the CB requirements needed, like the bandwidth & the number of critical flows

• Redundancy in the first & last links:

− Option A: Use dual-homed end nodes

▪ May keep the backbone’s bandwidth close to the same loading as before

▪ Cost = 1 extra link per critical end node & more CPU cycles to run 802.1CB for the critical flows only

− Option B: Use temporal redundancy single-home end nodes saving the cost of the extra links

▪ But the backbone’s bandwidth is still doubled and it also adds…

▪ Cost = More CPU cycles to run 802.1CB & duplicate frame transmission for the critical flows only

S-A S-B

S-Y S-Z

T-1

T-1

Page 17: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 6PUBLIC

Cost Factor – For Sensors & Their Flows

• Sensors can be very high bandwidth devices like 8+ gig/sec cameras

• Many sensors are needed too, 8 cameras is becoming a low number in a car

• Can this data even fit on a backbone today (even with reduced requirements)?

− With 6 camera @ 4 gig/sec requires 24 gig/sec! With redundancy that grows to 48 gig/sec.

− What year will these Ethernet PHYs & Switches be cost effectively available for Automotive?

• This bandwidth does not take into account the added data for Lidar, Radar, etc.

• And why double the bandwidth requirements of a sensor when the sensor itself is a single

point of failure?

− A camera “failure” is more likely due to dirt on the lens vs. a silicon or a wire failure!

− Why not add more cameras such that they overlap instead? Then CB is not needed for these flows

• Sensor Fusion merges data from many cameras & other sensor types to form a “picture”

− This process repeats continuously such that a Soft error is likely not critical and even many Hard

errors (like dirt) may allow for continued operation at reduced speeds (i.e., limp home)

Page 18: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 7PUBLIC

Cost Factor – For Actuators & Their Flows

• After Sensor Fusion, a decision is made on what to do

− Steer away from a problem or slow down, etc.

• Turning, braking, accelerating, etc. are Actuators

• Actuators:

− Are very low bandwidth devices

▪ Historically Actuators have been connected using CAN and LIN

▪ Thus doubling the bandwidth of these flows on the Ethernet backbone is totally feasible

▪ Even doubling the bandwidth, temporally, on a single link to the backbone is feasible

− Re-transmission (temporal redundancy) is quite often used today to overcome Soft errors

− Seamless redundancy greatly increases the likely hood that the 1st transmission is received

▪ Thus keeping the latency time for these Decision to Action messages low

Page 19: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 8PUBLIC

Summary & Conclusions

Page 20: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

1 9PUBLIC

Summary & Conclusion

• Hard, persistent errors are easier to see & plan around as noted in the examples

− The likely hood & impact of these errors (wire vs. software vs. silicon) is application dependent

• Soft, intermittent errors are harder to evaluate, but the impact of BER has been shown

− 1000BASE-T1 supports FEC making BER analysis much harder (possible future work?)

− The numbers presented for 100BASE-T1 are felt to be a good rule-of-thumb for 1000BASE-T1

• Redundancy is not free & the biggest cost today could be the extra backbone bandwidth

− Therefore, apply redundancy only to those flows that absolutely require it!

− It appears to be more practical to apply redundancy on Actuator as compared to Sensor flows

• End-to-end protection from talker to listener results in the highest integrity communication

• Dual-homing & temporal redundancy both offer very high robustness against soft errors

where multi-homing additionally protects against a single hard error to/from the ECU

• Using IEEE 802.1CB as shown allows development migration from “no redundancy” to

“best redundancy” with progressive steps (firmware updates) in between

− This allows re-use of existing ECUs without needing to rewrite/redesign everything in one go

Page 21: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

NXP, THE NXP LOGO AND NXP SECURE CONNECTIONS FOR A SMARTER WORLD ARE TRADEMARKS OF NXP B.V. ALL OTHER PRODUCT OR SERVICE NAMES ARE THE PROPERTY OF THEIR RESPECTIVE OWNERS. © 2020 NXP B.V.

Page 22: Practical Use Cases for Ethernet Redundancy...PUBLIC 8 Mixed Switches, Mixed Talker & Listener T-1 S-A L-1 S-B S-Y S-Z •Very similar Pros & Cons as previous slide •Only change

2 1PUBLIC

Verifying analytical model using Simulation -Based Fault-Injection

• Simulation with RTaW-Pegase, Bit Error Rate = 10-7

Solution Packet Loss rate by analysis Loss rate by simulation

Largest 2.44E-3 2.45E-3

Smallest 1.02E-4 1.04E-4

Largest 2.45E-3 2.47E-3

Smallest 1.02E-4 1.02E-4

Largest 1.33E-5 1.34E-5

Smallest 2.36E-8 2.14E-8

Largest 2.36E-5 2.35E-5

Smallest 4.19E-8 4.14E-8

Largest 1.23E-3 1.23E-3

Smallest 5.12E-5 5.12E-5

1

2

3

4

5

Simulation not suited for realistic BERs: e.g., one packet loss every 870 days

on average, for a 1ms stream (largest packet size) with BER=10-10


Recommended