With the sheer volume of data in today’s world, archive storage systems play a significant role in persisting the cold data. Due to stringent cost concerns, one popular design is to organize disks into groups and periodically switch them to be powered on for serving user requests. Scheduling thus becomes critical for both CapEx and performance. Unfortunately, field results indicate that existing schedulers can be often suboptimal. Our further analysis suggests that the main reason is the mismatch between the ever-changing workloads and the fixed set of coarsely-configured parameters in current heuristic-based schedulers. In this article, we propose MasterPlan , a reinforcement learning (RL) based scheduler for archive storage systems. By identifying the unique characteristics of archive storage service, we design a state space and reward function for the RL agent. MasterPlan includes a continuous action encoding approach to guarantee efficient exploration, and a meta adaptation module to extract features of workload series. Experiments show that MasterPlan can achieve 1.25× throughput, 2.16× 99 th latency and 1.47× power draw improvement compared to existing solutions.

Logs are ubiquitous digital footprints, playing an indispensable role in system diagnostics, security analysis, and performance optimization. The extraction of actionable insights from logs is critically dependent on the log parsing process, which converts raw logs into structured formats for downstream analysis. Yet, the complexities of contemporary systems and the dynamic nature of logs pose significant challenges to existing automatic parsing techniques. The emergence of Large Language Models (LLM) offers new horizons. With their expansive knowledge and contextual prowess, LLMs have been transformative across diverse applications. Building on this, we introduce LogParser-LLM, a novel log parser integrated with LLM capabilities. This union seamlessly blends semantic insights with statistical nuances, obviating the need for hyper-parameter tuning and labeled training data, while ensuring rapid adaptability through online parsing. Further deepening our exploration, we address the intricate challenge of parsing granularity, proposing a new metric and integrating human interactions to allow users to calibrate granularity to their specific needs. Our method's efficacy is empirically demonstrated through evaluations on the Loghub-2k and the large-scale LogPub benchmark. In evaluations on the LogPub benchmark, involving an average of 3.6 million logs per dataset across 14 datasets, our LogParser-LLM requires only 272.5 LLM invocations on average, achieving a 90.6% F1 score for grouping accuracy and an 81.1% for parsing accuracy. These results demonstrate the method's high efficiency and accuracy, outperforming current state-of-the-art log parsers, including pattern-based, neural network-based, and existing LLM-enhanced approaches.

Dimensionality reduction is an important step in knowledge discovery in databases. Intrinsic dimension indicates the number of variables necessary to describe a data set. Two methods, box-counting dimension and correlation dimension, are commonly used for intrinsic dimension estimation. However, the robustness of these two methods has not been rigorously studied. This paper demonstrates that correlation dimension is more robust with respect to data sample size. In addition, instead of using a user selected distance d, we propose a new approach to capture all log-log pairs of a data set to more precisely estimate the correlation dimension. Systematic experiments are conducted to study factors that influence the computation of correlation dimension, including sample size, the number of redundant variables, and the portion of log-log plot used for calculation. Experiments on real-world data sets confirm the effectiveness of intrinsic dimension estimation with our improved method. Furthermore, a new supervised dimensionality reduction method based on intrinsic dimension estimation was introduced and validated.

Feature selection is a critical preprocessing step in machine learning. It contributes to cost-effective model building and improvement of model prediction performance. Generally, a feature selection algorithm requires a dependency measure and a search strategy. Extant dependency measures are mostl y based on pair-wise correlation analysis, which cannot detect feature interaction. To overcome this problem, we developed a unified dependency criterion called inference correlation. The inference correlation between a set of predictor variables and a response variable can be efficiently calculated. The variables could be discrete, continuous, or mixed. Therefore, inference correlation can be applied to select features for both classification and regression problems. A feature selection algorithm using sequential floating forward search based on inference correlation is presented. Experiments of the algorithm on synthetic datasets and real-world problems confirm the effectiveness of the feature selection approach when compared to extant feature selection methods.