Struct AggregatedMetrics 

pub struct AggregatedMetrics {
    pub gpu_memory: HashMap<String, GpuMemMetrics>,
    pub cuda_latency: HashMap<u32, LatencyHistogram>,
    pub nccl_duration: HashMap<String, LatencyHistogram>,
    pub dataloader_wait: HashMap<u32, LatencyHistogram>,
    pub dist_sync: HashMap<u32, LatencyHistogram>,
    pub last_update_ms: u64,
}

Aggregated metrics computed from raw BPF events.

Fields

gpu_memory: HashMap<String, GpuMemMetrics>

cuda_latency: HashMap<u32, LatencyHistogram>

nccl_duration: HashMap<String, LatencyHistogram>

dataloader_wait: HashMap<u32, LatencyHistogram>

dist_sync: HashMap<u32, LatencyHistogram>

last_update_ms: u64

Timestamp (ms since epoch) of last update.

Implementations

impl AggregatedMetrics

pub fn record_gpu_mem(&mut self, pid: u32, gpu_id: u32, used: u64, peak: u64)

Record a GPU memory event.

pub fn record_cuda_latency(&mut self, pid: u32, latency_ns: u64)

Record a CUDA kernel launch latency.

pub fn record_nccl(&mut self, op: &str, duration_ns: u64)

Record an NCCL collective duration.

pub fn record_dataloader_wait(&mut self, pid: u32, wait_ns: u64)

Record a DataLoader wait time.

pub fn to_prometheus(&self) -> String

Format as Prometheus text exposition.

pub fn to_ws_snapshot(&self) -> Value

Build a JSON snapshot for WebSocket push to CLI.

Trait Implementations

impl Clone for AggregatedMetrics

fn clone(&self) -> AggregatedMetrics

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AggregatedMetrics

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for AggregatedMetrics

fn default() -> AggregatedMetrics

Returns the “default value” for a type.

impl Serialize for AggregatedMetrics

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.