G-release (Q1-Q2 2022)

Storage

1. NVMe/Vhost
   • Prototype
     • Implement creation of data queues
     • Rework Admin queue creation – move away from hardcoded implementation
     • Implement the minimum set of commands required to operate under linux
     • Implement Shadow Queue
     • Make sure works with Windows
   • Harden the prototype
     
     • Submit RFC patches to the mailing list once Prototype phase is ready
     • Address comments, work on cleaning up hacks
     • Run correctness tests, implement any missing bits an pieces
2. ZFS
   • First official support
     
     • Revise current storage health reporting in Eve OS
     • Add
       
       • Reporting multiple disks
       • S.M.A.R.T reporting
       • Zpool status errors
     • Enable reporting in the cloud
   • Zvol direct-io evaluation
     
     • Based on Delphix efforts presented on OpenZFS Dev Summit 2021
   • Async DMU / Async CoW
     
     • Deferring the reads so writes are not blocked
   • Online pool grow
     
     • If more disks added to the pool

Networking

Testing/Validation

Release Engineering

Security

D-release (Q4, 2020)

• Measured boot and Remote Attestation to the controller
• Encrypted disk vault keys sealed under the TPM PCR values
• End-to-end security even with content inspecting proxies
• Off-line support - verify that applications come up even if there is no connectivity to the controller
• Off-line support - user-setable reset/reboot time if connectivity is lost to controller (to recover from hung network adapters)
• App side-loading/pre-loading by being able to specify content trees independent of volume creation
• Use containerd as Content Addressible Storage for containers, VM and EVE images
• Verify device passthrough for primary video adapter
• Provide an API so Azure IoT edge workload can report status/metrics for the running modules via EVE to controller
• Non-virtio emulation of devices
• Reduce EVE memory footprint (single zedbox process for most microservices); apply C-group controls
• Experimental support for ZFS for the /persist filesystem including encryption

E-release (Target : Q1 2021)

• New EVE Logging Infrastructure (replace rsyslogd with simpler and more robust logging)
• Make IP geolocation more robust by moving to controller
  
• Enable deploying K3S workloads (including shared switch network instances to avoid NAT)
• Download EVE as OCI (one layer so far)
  
• Initial monitoring support for applications (crash, hang, ...)

F-release (Target : Q2 2021) TBD: add link to GH issue for each item

• Next generation storage architecture (design approved; in progress)
• Extract and report GPS coordinates to controller (if device has a cellular modem with GPS) (proposed)
• Incremental image download (rsync-like) (proposed)
• Better enforcement of volume size limits for the OCI-derived volumes (proposed - after storage rework)
• EVE self-monitor and log memory and disk usage increases - (first cut done - https://github.com/lf-edge/eve/pull/2023)
• EVE report more hardware counters? (e.g., disk/SSD S.M.A.R.T counters) (proposed)
• Combine eve, adam, and eden and just call it "eve"? Commands then looks like: "eve pod deploy ...", "eve generate server", ... and maybe after all the config is done, a command like "eve generate image" could be used to create an EVE-OS image for a specific board or hw arch (proposed)
• Expand EVE test coverage to improve overall stability: performance under stress; search for security vulnerabilities; automated code analysis; board farm daily and release testing (proposed)
• Controlling "VM bundles" (see below for a bigger scope around Local Controller) via "run profiles"
  • Controller feature to define the run-profiles, will allow customer to define the set of VMs that form their applications
  • Modifiable via MetaData APIs
  • User-owned APP responsibled for providing a local API endpoint on the box
• "VM bundle" VM version selection
• Hardening USB passthrough
• Hardening GPU passthrough
• Custom CA for HTTPs downloads
• GCP data store
• USB or SAN datastore

Future - to be planned

• Support multiple disks using different forms of RAID configurations (might use ZFS)
• Tailscale integration?
• Build EVE as OCI layers (reduces size of EVE update downloads)
• Local Controller
  • EVE toolkit (AKA MetaData server): read-only endpoints, protected by access control, physical access to the system required
    • Extending meta-data server API’s
    • provide local access to system metrics
    • provide a local access to change the run-profiles
  • HW Level metrics (IPMI-like): could be part of the above
• CPU affinity
  • Controlling "VM bundles"
• Infrastructure as code (Teraform plugin?)
• Snapshot/differential VM upgrades
• OCI observability
  • coredumps
  • crashkernel
• Sensor I/O virtualization
• Target upstream component for the next release
  • Alpine
  • Linux Kernel
  • Xen
  • Qemu
  • Golang