2025-10-09 (for public use)
1
In-Vehicle Networking Needs More Bandwidth
Courtesy: BMW
2025-10-09 (for public use)
2
Vehicles Increasingly are Defined by EE and SW
Courtesy: AMD
2025-10-09 (for public use)
3
Adopt & Adapt from CERN, Datacenter, …
At Team MLE we “borrow” from many different domains and make this applicable for others.
Example is Auto/TSN where we take networking technology from CERN and from Datacenter.
CERN White Rabbit
High Accuracy Time Synchronization
Low-Latency Datacenter Networking with
Ultra Ethernet
2025-10-09 (for public use)
4
Accuracy vs Precision
White Rabbit aims for sub-nanosecond accuracy and picoseconds precision
MLE “Light Rabbit”
2025-10-09 (for public use)
5
Datacenter: Reduce Tail-End Latency
Do Not Waste Compute Cycles in $1B AI Clusters
2025-10-09 (for public use)
6
Stanford University:
Homa
UltraEthernet Consortium:
UET
Tesla:
TTPoE
A New Wave of Transport Layer Protocols?
2025-10-09 (for public use)
7
QRP - Reliable, Rapid Request-Response Protocol
● Reliable = Acknowledge receipt
of Data
● Rapid = Connection-less and
makes use of priorities
● Request-Response = Message
based with Receiver-Diver
Congestion Control
● MLE runs HOMA Protocol on
Hardware Accelerator
2025-10-09 (for public use)
8
Effects of Hardware Acceleration
Linux-to-Linux via 10 GigE
FPGA-to-FPGA via 10 GigE
● Software (here Linux)
means ongoing memory
access, while modern
memory behave non-
deterministic (Caches!)
● Hardware (here FPGA) is
very deterministic,
stream-processing takes
10s of clock-cycles at 500
MHz
2025-10-09 (for public use)
9
Result: From Software to Silicon
Accelerating Automotive In-Vehicle Network Protocols for Zonal Architectures
Co-funded by MANNHEIM CeCaS
