Autoscaling Controller

Solution Architecture v1.0 · Data & AI Global Practice · 2026-09 · 37 views · Azure and open source

37 views 37 HTML views37 SVG37 draw.io Updated 2026-10-17
Architecture views

37 views, each in three formats.

Open a view to read it in full. Every SVG carries its diagram source inside it, so it opens in diagrams.net fully editable with no import step; the draw.io files are the same diagrams as plain source.

  1. 01
    System Context

    What sits inside the controller's boundary, what feeds it, and what it is allowed to change.

  2. 02
    High-Level Architecture

    Six stages: observe, decide, declare, converge, serve, explain — and one feedback edge that matters.

  3. 03
    The Control Loop

    The six steps of one 15 second interval, and the fact that it closes.

  4. 04
    Layered Architecture

    Seven layers, and the one rule about which way they may call.

  5. 05
    The Critical Design Decision

    The decision plane never calls a capacity API. The red arrow is the path the architecture forbids.

  6. 06
    Container Architecture

    Every deployable unit, and the regional state it is allowed to touch.

  7. 07
    Signal Plane

    Four classes of source, one common shape, and a gate that qualifies every sample.

  8. 08
    Decision Plane

    Sharded with a lease per shard, and what one shard does in one interval.

  9. 09
    Actuation Plane

    Converge actual towards desired, and survive a capacity API that is rate-limited and failure-prone.

  10. 10
    Integration Surface

    Five inbound interfaces, five outbound, and the rule that separates them.

  11. 11
    Authority and Arbitration

    A workload policy, a schedule, a forecast and a human can all target the same fleet. One of them wins, and it is decided in one place.

  12. 12
    Data Architecture

    Two stores whose loss is loss of control, and two whose loss is only loss of explanation.

  13. 13
    Target and Policy Model

    The registry: what a target is, what may constrain it, and what expires.

  14. 14
    Decision Journal Model

    What has to be recorded for "why did this target have this much capacity?" to be answerable without replaying a metric.

  15. 15
    Signal Lifecycle

    A sample from emission to decision input, and the three ways it can fail to get there.

  16. 16
    Durability and Retention

    The contract each store signs, side by side.

  17. 17
    Regional Partitioning

    Every store partitioned by region, and nothing on the control path crossing one.

  18. 18
    One Control Interval

    Tick to actuation call, with the two halves meeting only at the record.

  19. 19
    Which Signal Drives the Loop

    A declared hierarchy, and how conflicting signals are reconciled inside one interval.

  20. 20
    Policy Evaluation and Combination

    Three policy forms plus a forecast, one proposed number, then the clamps.

  21. 21
    Stability and Oscillation Control

    Five mechanisms in series, and the record that every suppression leaves behind.

  22. 22
    Scale-Out

    A sustained backlog breach, through the warm pool and past it, to registered capacity.

  23. 23
    Scale-In

    How removal is made safe, and the instance that is allowed to say no.

  24. 24
    Warm Pool Lifecycle

    Size, warm, hold, attach, measure, recycle — and what it costs to hold it.

  25. 25
    Four Classes of Target

    The same loop, four driving signals, four definitions of "serving".

  26. 26
    When the Capacity Is Not There

    Stockout, quota and scarcity are one problem; throttling is a different one.

  27. 27
    Deployment Architecture

    One region, three zones, and what surviving the loss of one of them actually means.

  28. 28
    Multi-Region Topology

    Two identical control planes that share intent and nothing else.

  29. 29
    CI/CD and Policy Promotion

    Two things ship here — the controller, and the rules it decides by — and only one of them can move a fleet.

  30. 30
    Observability

    Signal type by loop stage, and the alerts an operator can actually act on.

  31. 31
    Cost and the Price of Stability

    Attribution, efficiency, and the one number a policy is actually tuned against.

  32. 32
    The Operator's Loop

    Alert, ask the journal, buy time, propose a policy, prove it, promote or discard.

  33. 33
    Trust Zones

    Five zones, and every crossing authenticated and labelled.

  34. 34
    Identity Separation

    Three workload identities, federated, short-lived, and scoped to the capacity groups they manage.

  35. 35
    RBAC and Change Control

    Four roles, one change path, and the two things that need a second pair of eyes.

  36. 36
    What It Does When It Cannot See

    Fail static, then — only if declared — decay towards a schedule.

  37. 37
    Failure Modes

    Twelve classes assumed to occur, each with its named handling.

The package

Everything as it was delivered.

These files are served exactly as they were produced — the diagram pages keep their own house style because that is the artifact, not a rendering of it.