Milestones
List view
Deliver PoC documents
Overdue by 19 day(s)•Due by September 30, 2026•0/8 issues closedInvestigate options and create Architectural Design Reviews to take to the Technical Advisory Panel
Overdue by 1 month(s)•Due by August 31, 2026•0/5 issues closedThis milestone is to introduce the initial Technical Advisory Panel to the project
Overdue by 3 month(s)•Due by June 30, 2026•1/1 issues closed# MNeuData PoC EPIC ## High-level Aims As part of MNeuData, ILT have approved the idea of a proof-of-concept stage to explore some ideas and affirm design choices before moving to full implementation. ## Desired Outcomes 1. **Improve data rate estimates**, particularly for Endeavour instruments - Include feasibility of streaming all monitor data. 2. **Documentation on possible “bad data” scenarios** - What does the current ICP handle? 3. **Architecture Decision Records describing the following:** - Broker choice - Serialization format choice - Deployment topology choice, e.g. per target station / per facility - How/where within system the “bad data scenarios” will be handled - How non-streaming data can be integrated (e.g. IMAT cameras that only give access to a file) - Compute cluster technology choice - *(SH – Running/supporting K8S is frequently raised, so worth including here)* 4. **An end-to-end test system including the following:** - Multi-node broker cluster - Include deployment scripts/code/docs to demonstrate how the cluster would be administered - Multi-node compute cluster - Include deployment scripts/code/docs to demonstrate how the cluster would be administered - *(SH – Reminder to consider how other teams would deploy services here)* - Deployed in the chosen topology 5. **Cluster size estimates for costings** - Include commentary on retention times 6. **Networking requirements for chosen architecture** - Anthony Shuttle will need these for planning 7. **Document risks and risk mitigations** --- ## Demo Use Cases Once the above decisions have been made, demonstrate test scenarios. It most likely makes sense to use simulated data for testing. 1. **Disaster recovery** *(can be done on cloud?)* - Simulate operational error scenarios - Ensure understanding of recovery processes - Include documentation 2. **“Normal” operations** *(on real hardware?)* - Simulate real ISIS data rates - Do not need to produce “real” files that can be read, but demonstrate that the system...
Due by October 31, 2026•0/1 issues closed