Serverless platform driven CPU loadbalancing

Kernel-Level Dynamic Priority Scheduling for Containers

The use of containerized cloud platforms has now become inevitable for latency-sensitive applications; but resource contention between different workloads often creates the problem of noisy neighbors, which causes higher response latency and poor service quality. We present SCX_MUS (Scheduler Extensions for Multi-User Scheduling), a dynamic priority scheduling framework at the kernel level that enhances the CPU allocation to latency-sensitive containers running in Kubernetes environments. The proposed framework uses Linux sched_ext architecture along with the extension of eBPF to support dynamic cgroup-based scheduling through runtime awareness. A lightweight user space component continuously observes the workload of Kubernetes and adjusts the scheduling priorities of the containers without any need for recompiling the kernel or restarting containers. The experimental evaluation of the proposed framework was done with the help of Redis and stress-ng workloads in the same benchmarking condition compared to the native Linux Completely Fair Scheduler (CFS). The experimental results reveal a significant improvement in terms of latency reduction, enhanced throughput, efficient utilization of CPU resources, and stable performance of scheduling under resource contention.

Load-Aware Adaptive Scheduling (LAAS) System: A Real-Time CPU Scheduler for Minimal Operating System Kernels

Adaptive CPU scheduling that responds to dynamic workload characteristics remains an open challenge for minimal operating systems. Traditional schedulers employ static parameters unsuitable for heterogeneous workloads, while recent machine learning approaches incur overhead inappropriate for resource-constrained environments. This paper presents Load-Aware Adaptive Scheduling (LAAS), a novel approach that integrates real-time system load classification with entropy-inspired process scoring to enable dynamic scheduling decisions. The core innovation—a load-dependent scoring formula where weight values adapt based on system load class—represents the first application of entropy principles to CPU process selection logic in a minimal kernel. LAAS achieves adaptation without machine learning runtime overhead (sub-0.5% CPU vs. 1% for ML-based approaches) and is suitable for minimal monolithic kernels (approximately 1,500 LOC total implementation). Experimental evaluation on synthetic workloads demonstrates 31% reduction in average waiting time, 28% reduction in context switches, and 18% improvement in CPU utilization compared to standard Round-Robin scheduling, while achieving a Jain fairness index of 0.94 in our test scenarios. These results indicate that load-aware, entropy-inspired adaptive scheduling is practically feasible in severely resource-constrained kernel environments.

A Kubernetes Scheduler Plugin for Cluster-Wide Placement Optimisation

The default scheduler of Kubernetes, the state-of-the-art container orchestrator, uses fast, local placement decisions. Unfortunately, this design leads to resource fragmentation, reduced cluster usage, and overprovisioning. External solvers can compute global placement plans, but enforcing these plans in upstream clusters is hard. Kubernetes provides no native cross-node preemption, uncoordinated concurrent scheduling leads to inconsistencies, and replacing the default scheduler would sever deployments from upstream cycles. We present OPSche, an open-source Kubernetes Scheduling Framework plugin where external solvers can drive cluster-wide placement decisions in concert with the default scheduler. OPSche atomically validates and enforces solver-produced plans through coordinated framework hooks and supports three trigger modes: scheduling-failure, periodic, and stable-queue -- resp. triggered when a workload cannot be placed, at fixed time intervals, when the set of pending workloads stabilises. Each mode has a blocking variant for a finer tuning of placement quality, latency, and disruption. We pair OPSche with a constraint-based optimisation solver, showing its feasibility across a broad set of cluster configurations and reporting improvements of resource usage by up to 3.0% and scheduling latency by more than a second.

Future Generation Computer Systems

Performance evaluation of scheduling tasks in many-core systems utilizing processes and threads

This study assesses the scalability of process-based and thread-based schedulers for many-core shared-memory systems using a memory-intensive row-wise quick-sort workload on large three-dimensional tensors. The process-based evaluation considers bounded prolific, bounded collective, and three pipe-based producer-consumer schedulers: one-to-one, one-to-many, and many-to-many. These pipe schedulers dynamically stream task identifiers to worker processes, exchanging increased inter-process communication overhead for enhanced runtime load balancing and flexible chunk-based task dispatching. The thread-based evaluation examines static, dynamic, guided, chunk-based, chunk-stealing, adaptive chunk, and AIMD adaptive scheduling strategies. The AIMD scheduler employs an additive-increase multiplicative-decrease policy inspired by TCP congestion control, utilizing an exponentially weighted moving average (EWMA) of CPU utilization to regulate a contention window that limits the number of concurrently active chunks. The adaptive chunk scheduler further modifies chunk size based on observed per-thread execution speed. Experimental results on a 24-core x86-64 platform indicate that thread schedulers deliver the highest overall performance, with dynamic and guided scheduling yielding the most favorable practical outcomes. Among process schedulers, pipe-based designs demonstrate the strongest scalability, with one-to-one pipes excelling for smaller workloads and many-to-many pipes preferred for larger workloads. In summary, lightweight thread scheduling is optimal for shared-memory row sorting, while AIMD/adaptive scheduling and pipe-based process scheduling remain valuable for contention-aware execution, explicit inter-process coordination, and distributed-style heterogeneous workload management.

ARDA: I/O Scheduler for Heterogeneous Workloads Co-located on Ultra-low-latency SSDs