The success of large-scale pre-training paradigm, exemplified by Large Language Models (LLMs), has inspired the development of Time Series Foundation Models (TSFMs). However, their application to financial candlestick (K-line) data remains limited, often underperforming non-pre-trained architectures. Moreover, existing TSFMs often overlook crucial downstream tasks such as volatility prediction and synthetic data generation. To address these limitations, we propose Kronos, a unified, scalable pre-training framework tailored to financial K-line modeling. Kronos introduces a specialized tokenizer that discretizes continuous market information into token sequences, preserving both price dynamics and trade activity patterns. We pre-train Kronos using an autoregressive objective on a massive, multi-market corpus of over 12 billion K-line records from 45 global exchanges, enabling it to learn nuanced temporal and cross-asset representations. Kronos excels in a zero-shot setting across a diverse set of financial tasks. On benchmark datasets, Kronos boosts price series forecasting RankIC by 93% over the leading TSFM and 87% over the best non-pre-trained baseline. It also achieves a 9% lower MAE in volatility forecasting and a 22% improvement in generative fidelity for synthetic K-line sequences. These results establish Kronos as a robust, versatile foundation model for end-to-end financial time series analysis.

In-Network Aggregation (INA) offloads the gradient aggregation in distributed training (DT) onto programmable switches, where the switch memory could be allocated to jobs in either synchronous or statistical multiplexing mode. Statistical INA has advantages in switch memory utilization, control-plane simplicity, and management safety, but it faces the problem of cross-layer resource efficiency in job placement. This paper presents a job placement system NetPack for clusters with statistical INA, which aims to maximize the utilization of both computation and network resources. NetPack periodically batches and places jobs into the cluster. When placing a job, NetPack runs a steady state estimation algorithm to acquire the available resources in the cluster, heuristically values each server according to its available resources (GPU and bandwidth), and runs a dynamic programming algorithm to efficiently search for servers with the highest value for the job. Our prototype of NetPack and the experiments demonstrate that NetPack outperforms prior job placement methods by 45% in terms of average job completion time on production traces.

As the emergence of recently popular large language model, distributed training (DT) optimizes the performance via using different parallelization strategies, resource schedulers and advanced compression techniques. Meanwhile, a promising acceleration primitive, In-Network Aggregation (INA), offloads the gradient aggregation to programmable switches to further reduce the communication overhead using the switch memory. However, to the best of our knowledge, how to identify the performance bottleneck in real time remains challenging. In this paper, we build Argus, a performance bottleneck monitoring framework for INA. Argus implements an aggregation digest extracting mechanism for real-time monitoring of DT jobs at the multi-tenant, multi-rack clusters. Argus models aggregation to identify performance bottlenecks in aggregation, which assists the scheduler in deciding the resource allocation. Extensive evaluation and prototype implementation show that Argus provides real-time and packet-level aggregation monitoring for identifying bottlenecks in INA with minimal performance overhead.

We abstract a MIMO scenario in distributed data stream mining, where a stream of multiple items is mined by multiple nodes. We design a framework named MimoSketch for the MIMO-specific scenario, which improves the fundamental mining task of item frequency estimation. MimoSketch consists of an algorithm design and a policy to schedule items to nodes. MimoSketch's algorithm applies random counting to preserve a mathematically proven unbiasedness property, which makes it friendly to the aggregate query on multiple nodes; its memory layout is dynamically adaptive to the runtime item size distribution, which maximizes the estimation accuracy by storing more items. MimoSketch's scheduling policy balances items among nodes, avoiding nodes being overloaded or underloaded, which improves the overall mining accuracy. Our prototype and evaluation show that our algorithm can improve the item frequency estimation accuracy by an order of magnitude compared with the state-of-the-art solutions, and the scheduling policy further promotes the performance in MIMO scenarios.

Data stream processing is a crucial computation task in data mining applications. The rigid and fixed data structures in existing solutions limit their accuracy, throughput, and generality in measurement tasks. We propose Dynamic Hierarchical Sketch (DHS), a sketch-based hybrid solution targeting these properties. During the online stream processing, DHS hashes items to buckets and organizes cells in each bucket dynamically; the size of all cells in a bucket is adjusted adaptively to the actual size and distribution of flows. Thus, memory is efficiently used to precisely record elephant flows and cover more mice flows. Implementation and evaluation show that DHS achieves high accuracy, high throughput, and high generality on five measurement tasks: flow size estimation, flow size distribution estimation, heavy hitter detection, heavy changer detection, and entropy estimation.

The dynamic nature of software-defined networking requires frequent updates to the flow table in the data plane of switches. Therefore, the ternary content-addressable memory (TCAM) used in switches to match packet header fields against forwarding rules needs to support high rates of updates. Existing off-the-shelf switches update rules in batches for efficiency, but may suffer from forwarding inconsistencies during the batch update. In this paper, we design and evaluate a TCAM update optimization framework that can guarantee consistent forwarding during the entire update process while making use of a layered TCAM structure. Our approach is based on a modified-entry-first write-back strategy that significantly reduces the overhead from movements of TCAM entries. In addition, our approach detects reordering cases, which are handled using efficient solutions. Based on our evaluation results, we can reduce the cost of TCAM updates by 30%-88% compared to state-of-the-art techniques.

Wildcard rules are implemented in various important networking scenarios, including QoS, firewall, access control, and network traffic monitoring and analysis. However, there are cross-rule dependencies between wildcard rules, which both increase significant overhead and affect the semantic correctness of packet classification when caching rules. Considerable efforts have been made to mitigate the impacts of the dependency issue in rule caching, but it is still a bottleneck for cache systems. In this paper, we show how to give applications the flexibility of completely dependency-free wildcard rule caching by decoupling the cached rules and their dependent rules. Our FreeCache scheme has wide applicability to packet classification devices with wildcard rule caching. We validate the effectiveness of FreeCache through two respects: (1) Implementing various cache algorithms (e.g., LSTM) and cache replacement algorithms (e.g., ARC, LIRS) that are difficult to use in dependency-bound situations in the cache system with FreeCache. (2) Developing a prototype in a Software-Defined Network (SDN), where hybrid OpenFlow switches use TCAM as cache and RAM as auxiliary memory. Our experimental results reveal that FreeCache improves the cache performance by up to 60.88% in the offline scenario. FreeCache also offers the promise of applying any existing caching algorithms to wildcard rule caching while guaranteeing the properties of semantic correctness and equivalence.

Software-defined networking (SDN) provides flexible network control, which has enabled new, more complex network control mechanisms. These approaches impose high demands on flow updates in OpenFlow switches. Existing SDN controllers and firmware have been optimized for high-speed network updates, but specific switch implementations differ in their behavior. Previous optimization schemes for updates ignore the diversity of instruction types and switch behavior. In this paper, we present RuleTailor, an efficient, measurement-based optimization framework for SDN flow updates, to overcome these limitations. In contrast to other measurement-based optimization frameworks, RuleTailor uses new techniques, such as instruction type transformation, pseudo deletion, and match field distance, which can work in both the control plane and the data plane. Our evaluation shows that RuleTailor achieves a performance of fewer than 12 milliseconds per update in a content-addressable flow table with 1k entries, outperforming the state-of-the-art measurement-based framework, Tango, by a factor of 10.