on Operating Systems

PagePilot: Synergizing Heterogeneous Backend Devices with Reusability-Aware Page Offloading

Page swapping remains the dominant mechanism for extending physical memory in modern operating systems. Despite the emergence of new memory technologies, swap-based memory extension continues to be widely deployed due to its transparency and compatibility with diverse storage backends. However, current Linux swap management relies on a simple priority-based policy when multiple heterogeneous backends are available, and is oblivious to page reuse behavior. As a result, fast devices are often occupied by pages that are rarely reused, while frequently refaulted pages are placed on slow devices, leading to unnecessary performance degradation under memory pressure. We present PagePilot, a reuse-aware page offloading framework for heterogeneous swap systems. PagePilot persistently tracks refault behavior across eviction and refault cycles and uses average refault distance to guide backend selection. In addition, a background migration mechanism corrects misplacements to preserve fast-device capacity for pages with imminent reuse. We implement PagePilot in the Linux 6.3 kernel and evaluate it using representative data-serving and analytics workloads. Our results show that PagePilot improves application throughput and reduces page-fault handling latency under memory pressure, while substantially increasing the fraction of refaults served by fast backends.

Invisible Yet Dominant: Big Stalls of Kernel I/O Mechanisms in Cloud OLTP Databases

Most databases, including PostgreSQL, RocksDB, and recent AI KV-cache middleware, rely on buffered I/O, delegating write-back to the Linux kernel. On the distributed block storage standard in the cloud, this delegation inherits a hidden bottleneck: each device is drained by a single kernel flusher thread over a high-latency, shallow-queue path. When the drain falls behind, dirty throttling pauses write() system calls, and even reads that must evict dirty pages stall. These stalls are invisible to iostat and every standard counter. This poster observes the stall from inside the kernel, using the multi-volume data placement proposed in SteelDB as the experimental lever. eBPF probes on writeback and block tracepoints separate write-back by issuing context and count every throttle pause. Across three configurations with identical provisioned IOPS and bandwidth but 1, 2, and 4 devices, we show that adding drains, not bandwidth, cuts throttle pauses by 70%, reduces maximum transaction latency by 59%, and raises throughput by 23%.

Tierce: Observability-Driven Tiered Memory Management for Colocated Workloads

Characterizing Predictability–Latency Trade-offs of KV-Cache SSD Offloading in LMCache for LLM Serving Systems

KV-cache offload is widely used to stretch GPU memory for LLM serving, but its storage behavior has not been characterized at the block-device level. In this paper, we study LMCache through realworld multi-session workloads that span same/different context $\times$ same/different prompt, using over 100 stateless requests per workload. Our comparison is observational rather than factorial: it contrasts two realizable deployment snapshots—LMCache 0.3.0 with buffered I/O and the kernel page cache, and LMCache 0.3.16 with prefix-aware deduplication and O_DIRECT. Key findings define the paper. First, the legacy stack appears almost read-free at the SSD layer in all four workloads, but a flush_ram experiment shows that this is conditional on page-cache warmth: once the cache is evicted, the same path issues about 1 GB of sequential reads in 683 ms. Second, deduplication delivers a dramatic write reduction only for the exact-repeat workload: one set collapses from 1.62 GB to 11.75 MB of writes, while the other three sets remain in the 2–5 GB range. Third, O_DIRECT converts a hidden and bimodal read cost into an explicit and stable one: on one set, warm TTFT rises from 27 ms to 93 ms because each of the 99 warm requests re-reads the same ~12 MB partial chunk from SSD. The main conclusion is therefore not “newer is better,” but a system trade-off: predictability versus average-case latency.

NLock: A Scalable Lock for NUMA Architectures

The design of NLock enables an efficient local hand-over mechanism and in turn, achieves substantial performance advantages by co-optimizing lock design with modern NUMA architectures, and is evaluated against state-of-the-art locking mechanisms and concurrent data structures.

FreeToken: Efficient Edge-Native MoE Serving with Bandwidth-Adaptive Execution

Frontier open-weight models are increasingly available, but serving them still largely assumes datacenter infrastructure. We present FreeToken, an edge-native MoE serving system that treats a personal machine not as a small GPU, but as a unified, elastic inference platform. FreeToken co-designs the full serving stack, including model layout and loading, expert residency, CPU--GPU execution, agentic state reuse, and runtime memory management, around two realities of local AI: agent workloads continuously change their execution pattern, and edge hardware exposes heterogeneous resources whose balance differs from machine to machine. Rather than committing to a fixed offloading strategy, FreeToken continuously maps computation and model state onto the resources actually available. FreeToken supports more than 20 MoE models and real coding and tool-using agents across hardware ranging from an 8GB laptop GPU to a single workstation GPU. More importantly, it changes what these machines can practically serve, from a 35B model on a laptop to a 284B model on a gaming desktop and the 753B GLM-5.2 on a single workstation GPU. FreeToken turns open weights into deployable local software, making the machines users already own a practical platform for frontier-scale intelligence. We release the system at flashml.ai.