Purpose and Rationale
The Hypervisor Project follows after two successful workshops at the last two GENIVI All-Member-Meetings and investigates the wide scope of open- source and commercial hypervisor technologies, and addresses challenges in their use. Through collaboration between all vendors, experts and adopters of virtualization technology we can lower the barriers to successful product development. The project drives requirements, standardization for Hypervisor APIs, and other types of investigations to facilitate ECU consolidation, price reduction, and management of mixed-criticality in systems for improved security and functional safety.
Next Meeting
Tuesday, May 29, 10:00 AM CET
Agenda:
- Sub-topic definitions (continuation of last week's discussion)
- API standardization, more clarity on project starting point, outlook and dgoals
- Concrete steps for architecture / use-case proposal (KPIT) → document?
- Discuss Milestones, deliverable material.
Topic Introduction
- Informative and teaching everyone to a similar level of understanding
- Technical neutral information, no sales pitch
- Try to open questions for discussion
- Define what is the remaining problem
- Technology needs improvement?
- Uncertainties?
- Information/documentation/lacking understanding?
- ....
- Suggestion:
- Final slide could be a list of questions (to the workshop participants)
AMM Workshop Agenda (Topic name & Responsible)
This is a workshop session spanning several hours. We expect significant participation from all attendees. Each topic should have a short introduction (maximum 10 minutes) followed by interactive discussion. Individual topics should be introduced by a variety of Hypervisor vendors and other experts, such as OpenSynergy and Perseus.
Discussion topics will include:
- Workshop introduction and intention
- Introduced by: Gunnar (GENIVI), Sang-Bum (Perseus)
- History of HVs
- Introduction to XEN project
- by: Lars (Xen) (TBC by Sang-bum)
- Market Overview
- by:Franz Walkembach (SysGo)
- includes an intro to certification
- Requirements gathering
- HV vendors asking OEMs/adopters/customers/etc to clarify technical requirements
- Introduced by: Matti (Open Synergy)
- Performance comparison between open source software based hypervisors on ARM SoC
- HV design and implementation
- Virtualization for Multi-core, SoC peripheral hardware and special-purpose CPUs
- Introduced by:Artem (EPAM) + input on Samsung roadmap (Sang bum)
- Standardization of hypervisor APIs (virtio and friends)
- Introduced by: Open Synergy (virtio intro + what's needed?) + ?
- How can reliability & safety be quantified?
- Introduced by:
- - how to invite experts? TUV/similar?
- Audio system design with HVs
- Introduced by: Artem (EPAM)
- Discussing GENIVI AudioManager extensions
- Graphics/GPU Sharing (in relation to GSHA project)
- possibly show-and-tell multiple vendors?
- Introduced by: - reaching out to Renesas ?
- Health/Debugging/Analysis/Logging (in relation to SHDA project)
- Introduced by:
- (Cyber-)Security enhancements based on virtualization
- System design and working with HVs from users/system integrator's perspective.
- Introduced by:
- ? tier-1 or OEM leading discussion?
- Hardware integration. Communication issues between vendors?
AMM Workshop Agenda Topics (brainstorm and more details)
- Requirements gathering
- Requests from HV vendors for OEMs/adopters/customers/etc to clarify technical requirements,
e.g. on particular driver, performance, reliability, safety. - Central device driver management?
- ^^ Lead/intro by Sang-Bum
- i.e. HV+BSP supporting hardware register access arbitration.
- ...?
- Even possible without this? What would be design patterns to use if the HV is not giving the hardware virtualization you need.
- BSPs available vs. needed
- Unikernel support
- e.g. tested/demonstrated
- any special matching needed between HV and unikernel (para-virtualization) – likely not...
- (Maybe in particular, how is reliability & safety quantified?)
Boils down to: Several safety related parts brought together - how to design (and evaluate) the total system. - Footprint
- Memory usage
- # Lines of code (auditing/reliability/trust...)
- Training / informative materials?
- Vendor-independent
- From users/system integrator's perspective.
- What can be done today, what can't be done today
- Consequences of para virt or hardware support for virtualization, and related BSP design
- Connection between "management/sales" slides and the very detailed HV documentation. Knowledge required in the gap here.
- Cross-domain architectures, examples, recommendations
- (repeated from above) Boils down to: Several safety related parts brought together - how to design (and evaluate) the total system.
- (repeated from above) Consequences of para virt or hardware support for virtualization, and related BSP design
- Is design even possible without virtualized hardware? Design patterns to coordinate access if the HV is not giving the hardware virtualization you need?
- 12 factors for service-oriented platforms.
- Required technology demonstrations
- Market/Technology survey
- Requirements survey - Questionnaire (to OEMs)
- Results could be statistics/anonymized if needed.
- Start work on this questions list.
- Feature evaluation
- Compare features
- Separate BSPs - paravirt or hardware virt support?
- ... or simply list "what types of features can a Hypervisor have"
- Comparing types of HV / key characteristics?
- Graphics/GPU Sharing (in relation to GSHA project)
- What things are hardware dependent?
- How to share graphics buffers?
- Health/Debugging/Analysis/Logging (in relation to SHDA project
- Audio...
- Past approaches are known and can be discussed (Sang-Bum)
- e.g. paravirt devices
- audio mixing in operating system can handle it instead of paravirtualized drivers
- Complete architecture idea, e.g. with GENIVI AudioManager
- One hardware run Cluster
- How to minimize impact on the development process when using HVs
- Glossary/terms - does everyone have the same understanding of terms?
Virtual Device standardization
Common I/O devices for hypervisor guests with standardized features and interface, such that device drivers (and thereby systems) are more portable.
Advantages:
- Device drivers (for paravirtualization) for the (Linux*) kernel don't need to be maintained uniquely for different hypervisors
- Ability to move hypervisor guests between different hypervisor environments
*virtio supported by BSD, Windows, Fuchsia, and others
Extending this: Standardizing a contract/standard between guest and hypervisor. Compare the OCI initiatives for containers. Container runtimes → can we have standardized "hypervisor runtime environment" that allows a standards compliant virtual (guest) machine to run.
- Hypervisors can fulfil the specification (with local optimizations / advantages)
- Similarly, this specification is what guests can be engineered to.
Compare: Linux Device Tree – ability to discover and configure devices.
Project Planning
TODO:
- Introduction to project, current status and what needs to be done
- Break down into actions
- Resourcing - clarify group participation & time spent (roughly)
Automotive Hypervisor Compatibility Definition
- Collect from different sources.
- The AGL software defined connected car architecture (a.k.a "AGL virtualization paper")
Related introduction from OpenSynergy at All Member Meeting, Seoul, 2017. –> see PDF presentation
^^^ NOTE that is a snapshot of a living paper, (see Google Docs edit location).