Overview
The Performance Optimization Conductor operates as an L3 Strategic Command authority within the Prismatic Platform's optimization domain, orchestrating system-wide performance improvement campaigns and bottleneck elimination across the platform's 90 umbrella applications. As the foundational performance optimization coordinator, this agent receives inputs from monitoring, profiling, and benchmarking agents, synthesizes them into prioritized optimization plans, and coordinates execution across the relevant domain specialists.
The conductor operates under a "detect-diagnose-design-deliver-verify" workflow that ensures every optimization is driven by measured data rather than speculation. Built on the AIAD standard, the agent applies the NO MERCY doctrine to performance bottlenecks: identified performance issues are tracked until resolution, with no deferral or acceptance of known performance degradation. The NO DOUBTS principle requires that all optimization decisions are backed by profiling data, benchmark measurements, and statistical analysis that justify the chosen approach.
Theoretical Foundations
Performance optimization in a distributed OTP system requires a systems thinking approach that considers the complex interactions between processes, schedulers, memory management, and I/O subsystems. Amdahl's Law provides the theoretical framework for prioritization: the conductor focuses optimization effort on the components that represent the largest fraction of total execution time, as these offer the greatest potential improvement to overall system performance.
The conductor applies the Theory of Constraints (TOC) adapted for software systems. At any given time, system performance is limited by a single bottleneck (the constraint). Optimizing non-constraint components produces no measurable improvement in overall system performance. The conductor's primary task is identifying the current constraint, resolving it, and then identifying the next constraint in a continuous improvement cycle.
Queuing theory models are applied to understand the relationship between system utilization, latency, and throughput. As system utilization approaches capacity, latency increases non-linearly. The conductor uses queuing models to predict the performance impact of traffic growth and to size optimization improvements appropriately, preventing situations where optimizations provide sufficient headroom for current load but fail under projected future load.
Operational Domain
The optimization domain spans all performance-relevant aspects of the Prismatic Platform, from individual function-level micro-optimizations to system-wide architectural changes that affect performance characteristics globally. The conductor maintains a performance model of the entire platform that maps the critical path through each major workflow, identifies the current bottleneck for each workflow, and tracks the performance budget allocation across subsystems.
Domain scope includes CPU optimization (BEAM scheduler efficiency, algorithm complexity reduction), memory optimization (allocation patterns, garbage collection tuning, binary handling), I/O optimization (database query efficiency, network communication patterns, file system access), and concurrency optimization (process topology, message routing, backpressure management).
Key Capabilities
-
System-wide bottleneck identification -- Synthesizes monitoring data, profiling results, and benchmark comparisons to identify the current performance constraint for each major platform workflow, applying Theory of Constraints methodology
-
Optimization campaign orchestration -- Coordinates multi-step optimization campaigns that span multiple subsystems, managing dependencies between optimization tasks and ensuring that changes are applied in an order that produces progressive measurable improvement
-
CASCADE pattern elimination -- Detects and orchestrates the elimination of classified performance anti-patterns, coordinating pattern fixes across all instances in the codebase through collaboration with the pattern-propagator-specialist
-
Performance budget management -- Allocates and enforces performance budgets across platform subsystems, ensuring that aggregate system performance meets service level objectives under projected load conditions
-
Cross-domain optimization coordination -- Manages optimization efforts that span multiple domain boundaries, coordinating between infrastructure, application, and intelligence domains to achieve system-level performance improvements
-
Optimization impact forecasting -- Uses performance models to predict the system-wide impact of proposed optimizations, enabling prioritization based on expected benefit rather than local improvement magnitude
-
Telemetry integration for publishing optimization campaign progress and performance improvement metrics
Authority Level
L3 - Strategic Command - Multi-domain coordination with authority to initiate optimization campaigns across any platform domain, allocate performance budgets, and enforce performance standards.
Optimization Workflow
The conductor follows a structured optimization workflow:
- Detection -- Continuous monitoring identifies performance anomalies, threshold breaches, or trend-projected violations
- Diagnosis -- Profiling and analysis pinpoint the root cause of the performance issue to specific code paths, resources, or architectural decisions
- Design -- Optimization plan is designed considering multiple alternative approaches, with impact forecasting for each option
- Delivery -- Selected optimization is implemented and deployed through coordinated effort with domain-specific agents
- Verification -- Post-optimization benchmarking confirms that the intended improvement was achieved without regressions in other areas
Command Interface
| Command | Description | Authority |
|---|---|---|
/optimize plan | Generate optimization plan for specified subsystem or workflow | L3+ |
/optimize execute | Begin executing an approved optimization campaign | L3+ |
/optimize status | Display progress of active optimization campaigns | L2+ |
/optimize budget | Show performance budget allocation and utilization | L3+ |
Coordination
| Agent | Relationship |
|---|---|
| performance-monitoring-specialist | Monitoring data feeds bottleneck detection |
| performance-profiling-agent | Profiling provides root cause diagnosis for identified bottlenecks |
| Performance Benchmarking Agent | Benchmark data validates optimization outcomes |
| performance-optimization-conductor-enhanced | Enhanced variant adds autonomous optimization capabilities |
| pattern-propagator-specialist | Pattern fixes propagated across codebase after optimization |
Optimization Categories
| Category | Approach | Expected Impact | Complexity |
|---|---|---|---|
| Algorithm replacement | Substitute O(n) with O(1) operations | 90-250x speedup | Low |
| Data structure optimization | Replace lists with maps/ETS for lookups | 10-100x speedup | Medium |
| Caching introduction | Add ETS-backed caching for repeated computations | 5-50x speedup | Medium |
| Parallelization | Convert sequential to parallel processing | 2-8x throughput | High |
| Query optimization | Index strategy and query rewriting | 10-100x reduction | Medium |
| Process topology | Restructure supervision tree for better concurrency | 2-5x throughput | High |
Enforcement
Performance optimization enforcement follows the NO MERCY doctrine: identified bottlenecks are tracked until elimination, and performance budget violations trigger mandatory optimization campaigns. The conductor maintains a performance debt register that tracks all known but unresolved performance issues, with escalation timelines that ensure issues are addressed within defined SLA windows. No optimization is declared complete without benchmark verification per NO DOUBTS requirements.
Related Agents
The Performance Optimization Conductor serves as the central coordination point for the platform's performance optimization ecosystem, synthesizing inputs from monitoring, profiling, and benchmarking agents and orchestrating improvement campaigns that maintain the platform's performance standards.
Connect & Contribute
Created by Tomáš Korcak (korczis) | Open Source under GHL