Papers

Long-running LLM applications repeatedly send growing context, making prefix caching critical for reducing prefill cost. Yet prefix-cache behavior under agentic workloads remains poorly understood. We study production traces from two companies and evaluate 14 eviction algorithms across HBM-constrained and large memory-pool settings. Despite a large gap to Belady, sophisticated policies designed for traditional caches provide little benefit over LRU. The reason is structural: prefix reuse is dominated by the regular pacing of active sessions, making recency unusually predictive. Prefix caching nevertheless introduces new challenges, including heavy-tailed session footprints and highly variable miss costs as attention computation grows with sequence length. We introduce the compute-savings ratio and two offline oracles to quantify these effects. Our results show that effective prefix-cache management should retain recency as its foundation while selectively adding quick demotion for one-hit prefixes, compute-aware partial eviction for expensive misses, and capacity-dependent eviction granularity. We will release the traces and simulator to support future research.

GPU acceleration is now routine across robotics, but cloud-hosted GPU continuous integration (CI) runners are expensive, resulting in severe under-testing of GPU-accelerated code. We present pytest-gpu-proof, an open-source pytest plugin offering a practical middle ground. Tests can be run on a local machine, signed with a receipt of exactly what ran and what it produced, and integrated into standard CPU CI workflows (e.g., GitHub Actions). The tool is open source and on PyPI, and we are actively integrating it across our lab's software stack.

Today's digital healthcare services rely on various applications and functions that must be continuously accessible. Cloud computing enables global access to these services through public networks, which often also introduce increased latency, higher bandwidth consumption, and additional security risks compared to local operation. Edge computing mitigates these challenges by deploying cloud services closer to the end users and data sources, thereby improving resilience to network disruptions and reducing latency, bandwidth usage, and exposure to security threats. However, service deployment typically relies on the availability of centralized registry servers. Consequently, if the network connection or the registry server itself becomes unavailable, service deployment at the target edge node may fail. To address this, we propose a three-tier registry architecture to enhance deployment reliability and service availability, considering DockerHub as a remote public registry, an MEC-based off-premises registry as a remote private registry, and a LAN-based on-premises registry as a local private registry. The performance and efficiency are analyzed through measurements related to the estimated deployment time, inflicted network and computational load, and energy consumption, while the required nanoservices are deployed from different tiers. The experimental results indicate that alongside the improved tolerance to network disruptions, the local private and remote private registries also outperform the remote public registry in the deployment performance. The findings highlight the effectiveness of proximity-aware service distribution in improving the resilience, performance and efficiency of service deployment.

This paper presents a controlled comparison of exactly-once Kafka pipelines implemented with Apache Flink and Kafka Streams, two engines with different state-management architectures. The state management in Flink occurs through checkpoints in external storage, whereas Kafka Streams restores local state by replaying broker changelogs. The experiments evaluate the effects of the two designs on latency, resource consumption, configuration sensitivity, and recover from injected failures. The engines produced expected outputs across 50 correctness trials, five for each engine-workload combination (exact 95% CI 0.929-1.000). The latencies, however, showed differences when tested in 30 minute trials at a fixed rate of 100 events/sec. The measured ingestion to output interval was 4-6 ms for stateless workloads and 4.3-6.2 s for windowed workloads. The median p99 was 1.97-1.97s for stateless workloads and 5.68-9.92s for windowed respectively. Further, a sensitivity study showed that increasing the durability interval from 1,000 to 10,000 ms shifted the latency distributions in Kafka Streams and Flink between stages. Kafka Streams' W3 median p99 T2-T1 increased from 6,085 to 23,155 ms; Flink's T3-T2 increased from 732 to 8,702 ms. Failure injection experiments demonstrate that latency can interfere with incomplete processing behavior. The results show that exactly-once correctness is a necessary but insufficient measure of stream processing performance. Kafka Streams completed 0/5 stateful JVM-kill and 1/5 in each stateful local-volume-loss trials, while corresponding Flink cells completed 5/5. The cost, placement and magnitude depend on state management, architecture, configuration, and measurement location. Accordingly, evaluations of fault-tolerant streaming systems should holistically measure latency, completion, recovery behavior, and correctness.

Distributed data management systems deployed in untrusted environments rely on Byzantine Fault-Tolerant (BFT) consensus protocols to tolerate malicious failures. DAG-based BFT protocols improve throughput by letting validators disseminate transactions concurrently and by scaling execution across multiple workers. However, imbalances in workload or resource capacity can still degrade performance significantly. This paper presents Flamingo, a load-balancing protocol for certified DAG-based BFT protocols that addresses imbalance at both the ordering and execution layers. At the ordering layer, Flamingo periodically migrates client accounts away from overloaded validators, adapting to skewed submissions and heterogeneous validator capacity while preserving correctness under Byzantine faults, with migrations taking effect only through the committed log. At the execution layer, Flamingo redistributes committed transactions across executor workers using a deterministic, order-preserving scheduler that balances load and minimizes cross-worker data movement, without centralized coordination or costly distributed commit. Built on top of Narwhal and Tusk, our prototype shows that Flamingo recovers throughput and latency under workload skew, validator heterogeneity, and shifting hotspots, adds negligible overhead when the system is balanced, and needs load balancing in both layers, since resolving only one shifts the bottleneck to the other.

GPU robotics lacks the reusable numerical infrastructure of mature CPU stacks, instead relying on compiler frameworks that introduce overhead or repeatedly reimplementing numerical libraries. To address this, we introduce GLASS (GPU Linear Algebra Simple Subroutines), a header-only CUDA C++ library that provides thread-, warp-, block-, and NVIDIA-backed implementations of robotics-scale linear algebra and geometric computations under one composable device API. GLASS treats implementation choice, execution scope, and launch packing as architecture-specific placement decisions determined by offline measurement and resolved statically at compile time. This is critical as the best and worst placements differ by a median of 4.9x (max 81x), with 145 of 396 recommended placements changing between a Jetson AGX Orin and an RTX 5090, and 162 of 396 versus an AGX Xavier. These stakes are highest at the edge as GLASS's advantage over the best of PyTorch and JAX is as much as 73x on the Orin versus 12x on the RTX 5090. GLASS is released open source with independent numerical oracles and source-bound local-GPU test attestation. Finally, integrating GLASS with published robotics systems both exposed a pre-existing numerical bug and improved embedded runtimes by up to 1.5x.

Prefix caching is critical for efficient large language model (LLM) serving, particularly for agentic workloads that repeatedly invoke the model with a growing conversation and tool-use history. By reusing the key-value (KV) states of previously processed prefixes, prefix caching avoids redundant prefill computation. Its effectiveness, however, depends on retaining a sufficiently large set of KV cache states. Provisioning enough cache to preserve all historical KV states is prohibitively expensive and often unnecessary, whereas insufficient capacity can substantially degrade the cache hit rate. Determining the KV cache working set, defined as the minimum cache capacity required to achieve a target hit rate, is therefore essential for efficient cache provisioning and system design. We present KVSET, an online analyzer that estimates the KV cache working set of LLM serving workloads. KVSET uses the Mattson stack algorithm to efficiently estimate cache hit rates across a wide range of cache capacities. For each KV cache page, KVSET computes its LRU stack distance and compares it with the page number of each candidate capacity. This comparison determines whether the page would be a hit at each capacity without independently simulating every capacity configuration. KVSET therefore substantially reduces the computational and memory overhead of conventional capacity-by-capacity simulation and makes online working-set analysis practical. KVSET further determines the minimum cache capacity based on the maximum LRU depth among the prefix pages required to achieve the target hit rate. We validate KVSET using traces collected from production LLM workloads and show that its estimates closely match measurements from real cache deployments. The open-source implementation supports both online request processing and offline trace replay.

A Central Bank Digital Currency (CBDC) is central-bank money in digital form, held by the public. Leading designs have two limitations: conversion from bank deposits into CBDC can accelerate deposit flight, requiring safeguards, and the CBDC stays outside credit creation and monetary-policy operations. Here we present a CBDC architecture based on grassroots currencies that overcomes these limitations. The architecture has three components: (1) Money: sovereign grassroots coins, which are digital debts of one unit of fiat currency issued by the central bank, constituting a direct CBDC; (2) Credit and Liquidity: non-sovereign grassroots coins, which are digital debts of one unit of the same fiat currency, redeemable at par, that can be issued by any person, natural or legal, thus adding credit; and (3) Interest: grassroots bonds, sovereign and non-sovereign, adding maturity and thus interest, standard banking instruments, and the central bank's instruments of monetary policy. The central bank can therefore lend, absorb liquidity, set its rates and buy and sell securities in the coins and bonds the public holds, choosing the counterparties and terms of its credit operations, and without converting bank deposits into newly issued central bank money on demand. We prove that the arbitrage-free price of any non-sovereign grassroots coin whose issuer redeems it on demand is one unit of the fiat currency. The central bank can choose to deal with any counterparty, not just banks, and we argue that the central bank's interest rates on lending and bonds bound from above and below the corresponding interest rates of its counterparties. Sovereign and non-sovereign grassroots coins and bonds have been implemented and tested on a small scale.

Embodied reinforcement learning (RL) improves model capabilities with a pipeline of environment simulation, action generation, and model updates. These stages show heterogeneous CPU and GPU demands, making efficient resource utilization difficult. Recent systems overlap rollout (simulation and generation) with training for efficiency, but exclusive GPU allocation and synchronized barrier in rollout still leave substantial hardware resource waste. In this paper, we present EBRL, an asynchronous embodied RL training system with two core techniques. The asynchronous pipelined scheduler overlaps rollout and training, pipelines simulation and generation across environment groups, and carries out each environment independently, eliminating synchronization stalls. The fine-grained resource manager pools CPU cores and GPU streaming multiprocessors, and uses stage profiles and runtime feedback to adjust resource quotas and batch sizes to meet the shifting demands among stages. We implement EBRL on RLinf and evaluate it with four embodied policies and four simulation benchmarks across heterogeneous GPU testbeds. Experiments show that EBRL achieves 1.30-3.47 times the end-to-end rollout throughput and 2.5 times of training convergency compared to the SOTA embodied RL systems.

A major video platform runs on thousands of microservices, each request propagating a context so downstream work can be traced and governed. At handoffs outside instrumented paths, e.g., custom queues and callbacks, the payload continues but the context does not, and the request still succeeds under existing tests. Such breaks are silent and widespread: 673 of 1,133 services carried at least one. Backstitch, a specialized agentic system, repairs them using the surviving execution as its reference: replay determines whether a suspicious call is request-correlated, source analysis reaches the responsible handoff, a bounded change restores its contract, and the same replay validates the fix. Repairs restore the causal chain without disturbing the work it describes: breaks at 240 of the repaired calls fell from 90.46% to 4.69%, and over 112 days the fleet's break rate more than halved.

Algorithms in multi-agent systems such as federated learning, mobile robotic swarming, and consensus control can be designed and analyzed as distributed stochastic approximation algorithms. Such algorithms involve information exchanges between agents for various computations. The freshness of the information can be quantified using the Age of Information (AoI) metric. Consider robotic teams operating in highly obstructed geographical settings, such as subterranean or dense urban environments. Because of spatial disconnections, AoI has empirically been observed to be heavy-tailed with unbounded moments. However, most analyses assume AoI with bounded moments, creating a gap between theory and practice. To the best of our knowledge, ours is the first analysis under general heavy-tailed AoI with potentially infinite mean. We study the stability (almost sure boundedness of the distributed iterates) and convergence of multi-agent systems that are strictly dissipative in the scaling limit (system at “infinity”). Examples include most gradient-based and consensus algorithms under the Robbins-Monro step-size regime.

Building executable brain models is essential for moving neuroscience from description to mechanism and prediction. Human-brain-scale spiking simulation is constrained by highly irregular communication, multithreaded spike delivery, and the memory cost of sparse connectivity. We present CerebroSim, a scalable framework for whole-brain simulation. CerebroSim combines Delay-aware Spike Broadcast (DSB) for aggregated delay-aware communication, Race-free Synaptic Dynamics Computation (RSDC) for lock/atomic-free multithreaded delivery with HBM-aware optimization, and Sparse Synapse Storage Compression (3SC) for compact indexing with deterministic synapse regeneration. Using a model derived from magnetic resonance imaging and diffusion-weighted imaging, CerebroSim simulates 86 billion neurons and 100 trillion synapses on 18,432 nodes across 11.2 million cores of the LineShine Supercomputer, sustaining 24.44 PFlop/s, 91% weak-scaling efficiency, and 94% strong-scaling efficiency. This capability makes biologically constrained human-brain models practical for mechanistic studies of brain disorders and controlled in silico testing of intervention hypotheses.

Quantum cloud computing, delivered through the quantum-as-a-service (QaaS) model, provides access to quantum computing resources. However, applying uniform time-based pricing across fundamentally heterogeneous quantum resources significantly complicates task orchestration, particularly when addressing the tradeoff between execution costs and system performance. While heuristic methods rely on predefined scheduling rules, classical deep reinforcement learning (DRL) models may require more trainable parameters in this setting. Motivated by the potential of parameterised quantum circuits (PQCs) as compact function approximators, we propose QRLQ, a cost-delay-aware quantum cloud scheduling framework integrating PQCs with a dueling double deep Q-network (D3QN) to dynamically account for both cost and delay. Our simulation results show that QRLQ achieves lower mean cost and delay than the heuristic baselines, achieving a 5-11% lower mean cost relative to availability-based and rotation-based heuristics and reducing mean delay by 17% and 82% relative to the strongest and weakest heuristic baselines, respectively, while retaining execution fidelity within 2% of a fidelity-greedy policy. Compared with the classical DRL baseline, QRLQ achieves comparable scheduling performance while using 72% fewer trainable parameters. This work explores the feasibility of using QRL for task orchestration in quantum cloud environments and demonstrates its potential for cost-delay-aware quantum resource management.

Edge AI model deployment is a multi-stage engineering process involving model conversion, operator compatibility handling, runtime integration, and precision verification. While prior work has demonstrated agent-based automation for Qualcomm AI Runtime, the broader edge inference runtime ecosystem, including Intel OpenVINO, Rockchip RKNN, NVIDIA TensorRT, and ONNX Runtime, presents distinct toolchains and optimization strategies. This paper extends AIPC (AI Porting Conversion, an LLM agent-driven methodology for AI model deployment automation previously demonstrated on Qualcomm AI Runtime) to multi-runtime scenarios, proposing an LLM agent-driven approach for automated single-model-to-single-runtime deployment across heterogeneous inference backends, such as Intel OpenVINO, Rockchip RKNN, NVIDIA TensorRT, and ONNX Runtime. We decompose the edge AI deployment into standardized, verifiable stages, and inject runtime-specific domain knowledge into the agent execution flow through agent skills, auxiliary scripts, and staged verification loops. Using representative vision models, we demonstrate that agent-based deployment can complete the conversion from a PyTorch model to its executable inference, targeting OpenVINO for x86/NPU, RKNN for RK3588, TensorRT for NVIDIA GPU, and ONNX Runtime for Qualcomm NPU with a focus on FP16 precision deployment feasibility verification. The contributions of this paper primarily lie in providing multi-runtime deployment engineering practice experience, toolchain mapping analysis, a layout-adaptation and inference-replacement layer that removes manual transpose insertion from the agent's repair burden, and an empirical characterization of agent deviation behavior under structured knowledge injection, rather than large-scale systematic benchmarking or cross-runtime operator repair strategy comparison.

With the rising computational and monetary costs of training large language models (LLMs), checkpointing—periodically storing model states for recovery—becomes essential for fault tolerance. Conventional checkpointing entails a severe trade-off between checkpoint frequency (I/O overhead) and computational recovery (recovery time). State-of-the-art approaches mitigate this cost through pipelining checkpoint I/Os, differential checkpointing, or in-memory persistence, yet none leverage the distinct characteristics of LLM training dynamics, where model weight updates are non-uniformly distributed across transformer layers. This observation implies that saving all weights each time might not be efficient. Inspired by this observation, we present LayerCheck, a layer-wise adaptive checkpointing framework that selectively persists layers whose updates exceed a threshold. This design avoids periodic I/O bursts by distributing layer-wise checkpoint writes over time, resulting in smoother and more balanced I/O profiles. Upon recovery, LayerCheck reconstructs a mixed-timestamp composite model state by aggregating the most recently persisted versions of each layer together with their matching optimizer states. Under a bounded per-layer staleness guard, this introduces a controlled perturbation: under standard Adam assumptions it adds a bounded staleness term, and empirically the post-restart loss deviates from the failure-free trajectory by at most 0.54%. Empirical results on multiple open-source LLMs with different datasets further demonstrate that recovered models preserve the original convergence behavior and accuracy while substantially reducing checkpoint overheads. Specifically, LayerCheck achieves up to 22.6x reduction in total checkpoint size and 1.31x reduction in end-to-end training time compared to state-of-the-art systems, significantly lowering the cost of checkpointing.

Zeroth-order (ZO) optimization is an attractive option for memory-efficient LLM fine-tuning, but its fault tolerance remains underexplored. Unlike first-order training, ZO progress can be represented by lightweight seed-and-scalar step logs, yet naive log-only recovery still incurs replay cost that grows with training progress, and shortcut replay does not preserve the executed floating-point trajectory. We present ZOCheck, a fault-tolerant ZO training system that exploits this replayable structure through a CPU shadow process that continuously replays logged updates, materializes consistent recovery images off the GPU critical path, and persists them asynchronously. ZOCheck therefore combines non-blocking checkpointing during training with fast recovery from a near-current state. We also develop a cost model for choosing the snapshot policy under realistic failure rates. Experiments show that ZOCheck reduces checkpoint overhead by up to 219.7x and recovery latency by 1.55x on average compared with asynchronous full-state checkpointing, translating into up to 21.3x lower end-to-end wasted time across the evaluated failure rates, while preserving exact recovery behavior.

Phase unwrapping estimates the missing multiples of $2\pi$ in measured phase images. For large images, tiling limits the size of local reconstruction problems and enables parallel processing. Adaptive tiling could further reduce the number of local problems and boundaries by retaining large tiles where little refinement is needed. We investigate whether this reduction makes reconstruction faster. We compare complete reconstruction time and accuracy for a regular grid, quadtree, and kd-tree partitions. We also evaluate nine criteria for deciding where quadtree tiles should be subdivided, including residue count, fringe density, and measures of phase variation, at different tile sizes and budgets. In single-threaded experiments on a heterogeneous image dataset, optimized adaptive partitions use fewer tiles but remain slower than the optimized grid, and some reconstructions lose substantial accuracy. Stage measurements explain why: constructing the partition and solving larger retained tiles outweigh the savings at tile boundaries. The criterion comparison also shows that more refinement does not consistently improve accuracy. These results motivate evaluating adaptive partitions by the complete time needed to reach a chosen reconstruction accuracy, including whether limited refinement can provide a faster approximate result.

As serving capacity demand surpasses that of training, serving efficiency becomes increasingly important. Prefill-decode (P/D) disaggregation improves serving efficiency through specialization and isolation of the two phases. These benefits rest on a static partitioning. Phase demand, however, is not static. We observe that in a large LLM fleet the ratio of uncached input to output tokens has peak-to-mean ratios up to 4.7x at minute timescales, and that in a public agentic trace the hourly ratio spans a median 24.5x within a single day, while reassigning a replica takes tens of minutes. Agentic traffic sharpens the mismatch. Sizing each pool at its ninety-fifth percentile leaves up to 17% of cluster capacity unused; sizing below it converts the same imbalance into queueing and unrealized throughput. We present Crossflow, which makes this boundary elastic without changing node roles. Each decode node publishes a short-lived, revocable lease that bounds local-prefill compute, KV capacity, transfer work, and projected output. Across public and internal traces, Crossflow improves token throughput by 16.2-17.4% on geometric mean over static P/D, and by up to 43.4% at high load, while reducing mean TTFT at every evaluated point.

Multi-GPU servers have become the standard building block of modern data centers, providing aggregated capacity through high-bandwidth interconnects. At the same time, workloads such as LLM inference exhibit highly dynamic memory demands, which can cause one GPU to exhaust its local memory while others remain underutilized. This mismatch motivates a model of elastic resource sharing across GPUs. We present EMA, a memory sharing system that allows GPUs within a server to borrow and reclaim memory from each other, forming an elastic pool of capacity. EMA ensures performance transparency for both borrowers and lenders. For borrowers, prefetching hides remote access costs so that applications experience remote and local memory as indistinguishable in performance. For lenders, borrowed resources remain reclaimable on demand, guaranteeing that performance never falls below that of static partitioning. While our design focuses on memory, the same principle naturally extends to other GPU resources. Our evaluation shows that EMA improves individual user throughput by up to 52%, achieves 96% of the throughput of a system provisioned with 2X capacity, and maintains latency similar to the static local baseline.

We present a distributed quantum algorithm that $3$-colors cycles in $O(1)$ rounds, with high probability. It follows that all locally checkable labeling problems (LCLs) that have round complexity $O(\log^* n)$ in the classical LOCAL model can be solved in $O(1)$ rounds in the quantum-LOCAL model, with high probability; this includes problems such as maximal independent set and maximal matching in bounded-degree graphs. This presents the first natural examples of graph problems with an asymptotic distributed quantum advantage for the LOCAL model; all prior examples that separate LOCAL and quantum-LOCAL are artificial problems constructed merely for the sake of demonstrating quantum advantage.

Recent advances in large language models (LLMs) are driving the emergence of multi-modal and agentic services for mobile users through cloud and edge infrastructures, where long-context workloads pose daunting challenges for inference latency. Existing disaggregated LLM serving systems largely rely on hardware profiling, configuration enumeration, or heuristic scheduling, offering limited analytical guidance for cost-efficient resource allocation. In this paper, we propose SARA, a Service level objectives (SLOs)-Aware Resource Allocation framework for disaggregated agentic LLM serving systems, which maximizes goodput under a deployment cost constraint and a series of quantile-based SLO constraints. By capitalizing on queuing theory, we first model the prefill, KV cache transfer, and decode stages as an M/G/k queue, an M/G/1 queue, and a generalized birth-death process, respectively. The analysis reveals that the prefill and decode stages are dominantly limited by computational capacity and high-bandwidth memory (HBM) resources, respectively. With these mathematical models, we further derive tractable tail behaviors of different stage-wise service level metrics for both light- and heavy-tailed workloads. These characterizations explicitly map workload, model architecture, and hardware parameters to stage-wise SLO constraints and minimum resource requirements. Finally, we develop an effective resource allocation framework to maximize system goodput under limited cost budgets. Simulation and hardware results demonstrate that the proposed framework accurately predicts the stage-wise SLO with mean errors below 5%, and improves system goodput by 26.6% on average over state-of-the-art baseline methods under the same deployment cost.

Distributed Filesystems (DFS) are a crucial component of modern computing environments, and their performance is critical to the success of all the facilities that rely on them. However, predicting the DFS I/O performance solely based on the storage system hardware is not trivial. In this paper, we address this challenge by presenting an empirical method that tries to quantitatively assess how hardware configuration choices influence the performance of a DFS using Ceph as a case study. We investigate the influence of three hardware parameters-number of CPU cores, amount of RAM, and disk bandwidth. To control these variables, we relied on the Linux hotplug interface and Cgroups, avoiding additional software overhead. Our results reveal that for the analyzed workloads, decreasing hardware resources does not always yield proportional performance losses. This method offers practical insights for designing cost-effective distributed storage systems, remaining general enough to be applied to other filesystems.

Motivation: The Multiple Longest Common Subsequence (MLCS) problem for an arbitrary number of sequences is an NP-hard problem in sequence analysis. Existing exact algorithms based on dynamic programming or MLCS-DAG pruning rapidly exhaust memory as sequence lengths and set sizes grow, while heuristic and hyper-heuristic approaches compromise solution quality and require heavy parameter tuning, respectively. Results: This paper presents the ARP heuristic that, for a given primary sequence, performs three key actions: (i) Adding-$\Delta$s, which incrementally builds a solution by adding subsequences, called $\Delta$s, of the primary sequence to an initial solution; (ii) Replacing-Subsequences, which enhances the diversity of the solution set via targeted replacement of subsequences common to all solutions; and (iii) Prioritizing groups of characters from the primary sequence that are likely to appear together in longer common subsequences. ARP allows the user to configure the degree of Replacements and Prioritization actions for carrying out quality-runtime tradeoff. Empirical evaluations on both synthetic and biological sequence sets demonstrate that ARP's fastest configuration AOnly finds significantly longer common subsequences than the BNMAS classical heuristic and its aggressive configuration attains solution quality comparable to state-of-the-art hyper-heuristic (UB-HH) while running 1.1$\times$-1.7$\times$ faster.

Serverless platforms commonly rely on bundled, memory-centric configurations, where CPU capacity follows the specified memory size. Resource decoupling reduces this waste, but can create external fragmentation by producing diverse CPU-memory shapes that leave residual capacity stranded across nodes. Memory disaggregation and tiering can turn these resource holes into usable capacity by enabling elastic memory locality, where instances use different mixes of local and remote memory. In this paper, we present Memoryless, a fragmentation-aware resource manager that exposes memory locality as a serverless control-plane primitive. Memoryless jointly selects and places function variants that differ in their compute allocation and memory-locality ratio. This allows the scheduler to match instance shapes to fragmented node capacity while preserving SLOs. We implement Memoryless on top of Knative and evaluate it with trace-driven workloads. Compared to state-of-the-art serverless placement frameworks, Memoryless reduces node usage by up to 40% in steady state and 46% at peak, while keeping SLO violations within 3%. It also reduces CPU stranding by 44%, showing that elastic memory locality can convert fragmented node capacity into usable resources.

Real-time avatar services run long-lived, GPU-backed sessions that transform a continuous text stream into speech, facial motion, rendered video, and RTC media. Elastic operation repeatedly reassigns these sessions across Workers. A reassignment updates shared placement state while the previous GPU runtime and RTC connection remain alive, opening a gap between control-plane assignment and execution authority. A stateful controller serializes ownership by retaining per-session authority, placing that state on the controller's scaling and failover path. We present DHSched, a control plane that manages long-lived stateful sessions through stateless peer Dispatch replicas. DHSched externalizes session ownership as a generation identified by (WorkerID, epoch). Dispatch conditionally commits each generation. Infer-Controller claims and revalidates the same generation before releasing state-advancing input. A stale runtime loses execution authority even if it remains alive. DHSched ranks candidates from periodically refreshed load views while keeping capacity admission authoritative. Recovery advances the same generation path via source-epoch guards and source-independent reconstruction. DHSched runs in our production avatar-serving stack, which reached 49,987 concurrent live sessions. On 10 June 2026, it completed 9,927 session migrations with no observed dual-owner conflicts or capacity overshoots. At the production peak, CreateLive latency was 98.3 ms at P99 and PushText latency was 16.5 ms at P99. Across 100,000 randomized ownership runs, assignment-time, claim-only, and WorkerID-only protocols permit stale execution, while DHSched records no dual-owner or stale-generation release.

Optical Circuit Switching (OCS) offers high bandwidth density and energy efficiency for LLM training, but incurs a non-negligible reconfiguration delay. Prior work typically schedules optical circuit switches independently of compute, using aggregate traffic demand to determine which circuits to provision and when. We argue that this separation creates a fundamental inefficiency: reconfigurations that ignore the compute timeline can stall communication, resulting in low circuit utilization and large buffer requirements. In this paper, we present Flux, a scheduler that optimally schedules optical circuit switches based on the structure of the entire workload. Flux remains effective across a wide range of switching speeds by reusing circuits and amortizing reconfiguration delay behind compute and communication. We show that Flux reduces training iteration time by up to $10\times$ and peak NIC buffer requirements by more than three orders of magnitude compared to traditional periodic schedulers.

DAG-based consensus protocols can achieve great throughput and the optimal three-message-delay limit for n = 3f+1 consensus. While two-delay protocols exist, they pay with reduced resilience (requiring 5f+1-style committees) or rely on fallbacks that sacrifice the DAG's high throughput. This paper introduces Hydrozoan, the first DAG protocol with a dual commit path under a hybrid fault model of f Byzantine and c crashed validators, on n = 3f+c+2p+1 validators. Leaders commit in two message delays whenever at most p validators are faulty, and in three otherwise, with no extra messages, no view changes, and multiple leaders per round. Both paths are evaluated on the same DAG, using a novel graded indirect rule to reconcile them so that every honest validator reaches the same decision. We show that under geo-distributed conditions, which path is faster is a property of geography rather than the protocol, as rounds reaching a remote region cost far more than those that do not. The (f, c, p) knobs place the fast quorum where the deployment requires it, allowing a commit in two message delays. If misconfigured, Hydrozoan can still commit in three message delays: Hydrozoan commits on whichever path fires first. We also present Optimal-Hydrozoan, a variant that tolerates one more fault on the fast path, the first construction to match the known lower bound. The safety and liveness of both protocols are machine-checked in Lean 4. Our geo-distributed evaluation shows that Hydrozoan matches Mysticeti's throughput, commits ~25% faster when the fast quorum fits fast regions, and falls back to three message delays when it does not or past p faults, where existing two-delay protocols stall.

The timeliness and accuracy of edge-based video analytics can be hindered by drastic reductions in available bandwidth (ABW) at last-mile edge routers, causing prolonged queuing delays. This work proposes Lizard, a system that leverages video-content-aware packet discarding to mitigate the negative effects of drastic ABW degradation that may frequently occur at a last-mile edge router by judiciously discarding packets that contain frame blocks less important to the analytics at the destination. To achieve this, we first devise a frame-block-aware RTP header extension to effectively decouple packet dependencies to encode frame blocks. Second, Lizard uses a priority-based feedback mechanism that dynamically evaluates packet priorities based on relative accuracy impacts. Third, we develop an adaptive phase-transition-based packet discarding strategy at the router to discard packets that represent unimportant blocks. Our evaluation of Lizard shows improvements over existing methods are substantial: 53.2% reduction in latency and 27.1% increase in analysis accuracy.

Large-model parameters have grown beyond the capacity of a single xPU, dispersing across multiple xPUs spanning distinct host domains, where xPU-to-xPU communication dominates overall system efficiency. Existing scale-up interconnect remains inadequate: network-based solutions built on Ethernet–such as RoCE (RDMA over Converged Ethernet)–introduce specific message-semantics and protocol-stack characteristics, and rely on fragmented per-host addressing, while conventional host-based fabrics are confined to a single host domain and lack cross-host unified addressing. This paper presents Co-Fabric, a bus-based interconnect that, unlike conventional bus designs, breaks host-domain boundaries to deliver unified xPU interconnection for scale-up superpods. Co-Fabric makes three contributions: a streamlined four-layer protocol stack achieving nanosecond-scale processing latency with native reliability; a cross-domain scaling and P2P mechanism that routes using port identifiers embedded in the packet header; and a unified address space built on shadow-device auto-enumeration. On a 64-xPU 3D-Mesh system, Co-Fabric cuts inter-node communication latency by over 50% and improves bandwidth by 2-5x over RoCE, accelerating DeepSeek R1 inference by 30%-80%. Moreover, since its streamlined four-layer protocol stack and higher data-communication efficiency reduce protocol and processing overhead relative to the Ethernet-based RoCE stack, Co-Fabric cuts the cost and power of the interconnect itself by up to 80% and 5%, respectively. These results demonstrate Co-Fabric's advantage for AI computing centers.

Automatic failure recovery is another difficult area of DevOps deployments due to the complexity of the system, high workload and the limitations of the traditional rules-based or manual approach. This study proposes a hybrid model integrating deterministic, rule-based recovery with the assistance of machine learning by fault prediction to automatically monitor failures, identify, classify and recover failures in real time. To test the architecture, a distributed log dataset of 100,000 records was used for models such as Decision Tree, Random Forest, Logistic Regression, LightGBM, Autoencoder with BiLSTM. The Decision Tree had a strong performance in defect identification with an F1 score of 80.8%, recall of 80.0%, precision of 81.6%, and accuracy of 89.4%. An impressive 83.3% success rate was achieved by the automated recovery measures, resulting in a 94.6% decrease in mean time to recovery (MTTR), a 95.4% reduction in downtime, and annual savings of 7,462. The combination of explainable rule-based logic and adaptive AI prediction in the framework means it is more resilient, operationally efficient, and scalable, and an effective, autonomous approach to DevOps environments evaluated under controlled experimental conditions.