�ݺ�ߣ

2017-05-10
Alexander Much
“Legacy” Software
Competence Centre Systems Engineering | Much | 2017-05-10 | Gate4SPIC | Public | © Elektrobit Automotive GmbH 2017.
All rights reserved, also regarding any disposal, exploitation, reproduction, editing, distribution, as well as in the event of applications for industrial property rights.

2Competence Centre Systems Engineering | Much | 2017-05-10 | Gate4SPICE | Public | © Elektrobit Automotive GmbH 2017.
“Legacy” from Webster’s dictionary:
1. a gift by will especially of money or other personal property : BEQUEST She left us a legacy of a million dollars.
2. something transmitted by or received from an ancestor or predecessor or from the past the legacy of the ancient philosophers The
war left a legacy of pain and suffering.
„Legacy System“ from wikipedia (emphasis added):
In computing, a legacy system is an old method, technology, computer system, or application program, "of, relating to, or being a
previous or outdated computer system.“
Often a pejorative term, referencing a system as "legacy" means that it paved the way for the standards that would follow it.
This can also imply that the system is out of date or in need of replacement.
„Legacy“?
Gate4SPICE: Legacy Software

• Business models will change
• The SOP is not the central mile-stone anymore
• Being able to operate and maintain the system
is the key factor for business as well as quality
• New software eco-systems, SW platforms
• Security >> safety!
Consequences for quality and for processes
Weak points:
• PLC-like process models
• Missing service and operations in Automotive
SPICE (-> SPICE for DevOps?)
DevOps, OTA: from product to service
Future
Sensor
Actuator
Standard
ECU
Base Function
Services &
Operation
Updates
Critical Update
Content
Critical Update
@SOP
Features
OTA:
Over
Lifetime
Operator,
Mobility-Provider
Backend
InfrastructureAnalysis,
Data
collection
& Fleet
Management

Driving Forces behind OTA
Analysis,
Data
collection
& Fleet
Management
Future
Driving not possible when
data connection missing
Sensor
Actuator
Standard
ECU
Base Function
Services &
Operation
Updates
Critical Update
Content
Critical Update
@SOP
Features
OTA:
Over
Lifetime
Operator,
Mobility-Provider
(OEM or a service provider
without own vehicle
production like UBER)
Backend
Infrastructure
€€€
€
€
€
€
€€
Service Provider
(new services)
??

• Software in essence is the ”design of a machine abstracted from its physical
realization”1.
• The most complex systems ever built are all software systems.
• The design of software needs to be controlled. There is no process variability
in the manufacturing. We only have the construction.
• Software is often changed and adapted?  less rigor in control?
• The FDA: ”[i]n fact, the opposite is true. Because of its complexity the
development process for software should be even more tightly controlled
than for hardware”.2
Why software is special
Software
General-Purpose
Computer
Special-Purpose
Machine+ =
1Nancy Leveson, Engineering a Safer World:
Systems Thinking Applied to Safety, MIT Press, 2011
(free at https://mitpress.mit.edu/books/engineering-safer-world)
2CDRH, General principles of software validation.
FDA, 2002.
Controlling software evolution is key
A complex software system is planned from
maintaining it, not from creating it
 „Maintaining maintainability“

„Legacy Software“ comes in different forms:
1. You created a system without enough project or technical
control (or understanding)
2. You are faced with maintaining or operating a system that
has been created only with the SOP in mind
3. You „inherit“ a system, e.g. you bought another company
Such a software system is intangible:
• You don‘t know enough about it
• Process documents may be inaccurate or plainly false
Handling „Legacy“ Software in Practice
What to do?
• Don‘t believe process documents:
The code is in the vehicle, not the documents
• Build hypotheses and test them (documents + code):
– Top-down by testing (in the sense of learning)
– Buttom-up by architectural reconstruction
• Both at the same time!
• See e.g.:
„If your map doesn‘t match the terrain, it‘s not the terrain that‘s wrong“.

In file.h:
#define LOOP_END 0
In file.c:
int i;
for (i = MAX; i >= LOOP_END; i--)
{
/* … */
}
Beware: a „MISRA-fication“ in practice
• A (stupid) MISRA checker is used
• Warns about a signed literal, should be suffixed with ‚U‘
• Question: what happens if ‚U‘ is appended to LOOP_END?
• Remark: this happens in completely different files.
• Answer: usual arithmetic conversions turn the signed integer ‚i‘
into an unsigned integer, cannot be negative
 an endless loop

Background:
• AUTOSAR started ca. 15 years ago
• Millions of units on the road
• From QM to ASIL-D
• Complexity of the standard is increased rapidly with each
release
 ca 500 kLOC + > 20.000 configuration parameters
• OEM and Tier-1 quality demands evolved
• Standards evolved
-> anyone remembers MISRA 1998?
Example: AUTOSAR basic software
Approach:
• Strong focus on product quality instead of process quality
 everything that helps product quality helps the processes
• Daily measurement and build
– Code metrics
– V&V metrics
– Tracing metrics
• Permanent re-factoring, metrics are never violated
• Incremental approach:
– Each change-set is „complete“
– State of a component is „ok“ before and after the change
• Build number of the night build from today: 155597

Example: Process Metrics
Id Description Prototype Development Mass Production
PercReqImpl Percentage of implemented reqs n/a >= 0.85 == 1.0
PercReqTst Percentage of tested reqs n/a >= 0.75 == 1.0
NrReqNotImpl
Reqs that are not or partially
implemented
n/a n/a == 0
NrReqNotTst Reqs that are not or partially not tested n/a n/a == 0
NrTstNoReq
Specified tests that do not correctly trace
to a req
== 0 == 0 == 0
NrSpecNoReq
Sobjects that falsely claim coverage for a
req
== 0 == 0 == 0
NrTstNoSpec Implemented but not specified tests == 0 == 0 == 0
NrTstNotImpl Specified but not implemented tests n/a n/a == 0
NrRuleViolations
Number of static requirements rule
violations
n/a n/a == 0

Trends and Impacts on EE Architecture
Clean Mobility
Electrification
Safe Mobility
Autonomous Driving
Intelligent Mobility
Connected Car
Trends Impact on EE SW Architecture Impact on Economy
HW Complexity Decrease
Safety & Security Needs
Connectivity
SW Complexity Increase
System Availability
Embedded Smart Device
Security Needs
Lower barrier for new OEMs
focusing on e-cars,
Enabler for mobility.
Growing complexity,
Change in architecture and
value chain,
Entry point for IT companies.
New user centric mobility
services and business models,
change in value chain

The „system“ in the future
UI
Computing
Cluster
Central
Computing
Cluster
Smart Antenna
Gateway
Remote IO,
Actors, Sensors
Smart
Sensors
Smart
Sensors Steering
Braking Battery
Engine
Back-end
System
Gigabit
Ethernet

�ݺ�ߣ

2017-05-10 Gate4SPICE: "Legacy Software"

More Related Content

2017-05-10 Gate4SPICE: "Legacy Software"