filter refactor finalizing

Signed-off-by: Nir Rozenbaum <[email protected]>

Author: nirrozenbaum

pkg/epp/scheduling/plugins/filter/least_kvcache_filter.go

@@ -36,7 +36,6 @@ func NewLeastKVCacheFilter() *LeastKVCacheFilter {
 // The intuition is that if there are multiple pods that share similar KV cache in the low range, we
 // should consider them all instead of the absolute minimum one. This worked better than picking the
 // least one as it gives more choices for the next filter, which on aggregate gave better results.
-// TODO: Compare this strategy with other strategies such as top K.
 type LeastKVCacheFilter struct{}
 
 // Name returns the name of the filter.

pkg/epp/scheduling/plugins/filter/least_queue_filter.go

@@ -37,7 +37,6 @@ func NewLeastQueueFilter() *LeastQueueFilter {
 // we should consider them all instead of the absolute minimum one. This worked better than picking
 // the least one as it gives more choices for the next filter, which on aggregate gave better
 // results.
-// TODO: Compare this strategy with other strategies such as top K.
 type LeastQueueFilter struct{}
 
 // Name returns the name of the filter.

pkg/epp/scheduling/plugins/filter/low_queue_filter.go

@@ -27,11 +27,15 @@ var _ plugins.Filter = &LowQueueFilter{}
 
 // NewLowQueueFilter returns a new LowQueueFilter.
 func NewLowQueueFilter() *LowQueueFilter {
-	return &LowQueueFilter{}
+	return &LowQueueFilter{
+		queueingThresholdLoRA: config.Conf.QueueingThresholdLoRA,
+	}
 }
 
 // LowQueueFilter returns pods that their waiting queue size is less than a configured threshold
-type LowQueueFilter struct{}
+type LowQueueFilter struct {
+	queueingThresholdLoRA int
+}
 
 // Name returns the name of the filter.
 func (f *LowQueueFilter) Name() string {
@@ -43,7 +47,7 @@ func (f *LowQueueFilter) Filter(ctx *types.SchedulingContext, pods []types.Pod)
 	filteredPods := []types.Pod{}
 
 	for _, pod := range pods {
-		if pod.GetMetrics().WaitingQueueSize <= config.Conf.QueueingThresholdLoRA {
+		if pod.GetMetrics().WaitingQueueSize <= f.queueingThresholdLoRA {
 			filteredPods = append(filteredPods, pod)
 		}
 	}

pkg/epp/scheduling/scheduler.go

@@ -66,7 +66,7 @@ func NewScheduler(datastore Datastore) *Scheduler {
 
 	defaultConfig := &SchedulerConfig{
 		preSchedulePlugins:  []plugins.PreSchedule{},
-		filters:             []plugins.Filter{&filter.SheddableRequestFilter{}, lowLatencyFilter},
+		filters:             []plugins.Filter{filter.NewSheddableRequestFilter(), lowLatencyFilter},
 		scorers:             map[plugins.Scorer]int{},
 		picker:              &picker.RandomPicker{},
 		postSchedulePlugins: []plugins.PostSchedule{},