Knative　is　a　popular　Kubernetes-based　platform　for　managing　serverless　workloads　with　two　main　components　Eventing　and　Serving.　The　Serving　primitive　helps　to　deploy　serverless　apps　as　Knative　services　and　automatically　scale　them,　even　down　to　zero.　However,　the　serving　module　uses　a　moving　average　method　to　calculate　the　number　of　pods,　that　calculated　based　on　past　data　and　not　good　at　accounting　for　future　changes.　In　this　paper,　we　use　double　exponential　smoothing　to　optimize　the　calculation　of　the　number　of　pods.　Preliminary　experiments　show　that　the　results　are　better　than　the　moving　average.