With advancements in machine learning (ML) technology and the deployment of large ML-as-a-Service (MLaaS) clouds, accurately understanding request behaviors in an MLaaS cloud platform is paramount for resource scheduling and optimization. This paper sheds light on the correlation of request arrivals in a representative and dynamic MLaaS workload – Alibaba PAI (an ML platform for artificial intelligence). For requests in the PAI workloads at the job, task, instance, and machine levels, our burstiness diagnosis reveals that the request arrival processes at all levels are significantly bursty. Additionally, our Gaussianity test indicates that the bursty activities in PAI consistently appear to be non-Gaussian. Our findings show that there exists a certain degree of correlation between request arrivals at each level over long-term time scales. Moreover, we reveal the self-similar nature of request activities in the various-level wild MLaaS workloads on Alibaba PAI through visual evidence, the auto-correlation structure of the aggregated process of request sequences, and Hurst parameter estimates. Furthermore, we implement a versatile workload synthetic model to synthesize request series based on the inputs measured from the PAI trace. Experimental results demonstrate that our model outperforms typical self-similar workload models, and can improve accuracy by up to 99% compared to them.

The design and performance optimization of flash-based storage subsystems are crucial for improving the system performance of Android-based smartphones. However, it highly relies on wisdom derived from mobile storage workload studies of smartphone applications. From the temporal perspective, our burstiness diagnosis reveals that the arrival processes of I/O requests in 33 smartphone applications, are significantly bursty, especially for read requests. This article studies the correlation of inter-arrival times of read and write requests, and compares the correlations for read and write requests in the four same types of mobile applications. We first observe that read requests of 85% of applications and write requests of 88% of applications present a certain degree of correlation over a longer time range. Then, we further conduct Hurst parameter estimation for these mobile applications with mature statistical tools. All estimated Hurst parameters are larger than 0.5, confirming the existence of self-similarity in a majority of mobile application workloads. Finally, we deploy a flexible I/O request generator for smartphone applications based on the parameters measured from actual traces. Experimental results show that the proposed generator can accurately generate request sequences for various mobile applications and more faithfully characterize the heavy-tail properties of I/O activities in mobile storage workloads than traditional models.

As cloud platforms become increasingly popular, accurately understanding I/O behaviors in modern cloud storage is of paramount importance for system design and optimization. This paper sheds new light on the correlation of inter-arrival times of both read and write requests at the block level in four representative cloud storage workloads – AliCloud, Systor’17, MSRC and FIU. Our study reveals that I/O arrivals at the block level are very complex in modern cloud storage. There is a certain degree of correlation in the long-term timescale for request arrival intervals in AliCloud and Systor’17_read. Request arrival intervals in MSRC, FIU and Systor’17_write, however, are almost uncorrelated. The Gaussianity test confirms that I/O burstiness appears to be Gaussian in AliCloud_write and Systor’17_read, but the burstiness is non-Gaussian in other workloads. Importantly, we unfold the existence of self-similarity in cloud storage workloads with a certain degree of correlations, via visual evidence, the autocorrelation structure of the aggregated process of I/O request sequences, and Hurst parameter estimates. We further design an alpha-stable workload model for synthetic I/O generation, and the experimental results demonstrate that our model has an edge over conventional models in terms of accurately emulating I/O burstiness.

One of the challenging issues in performance evaluation of parallel storage systems through synthetic-trace-driven simulation is to accurately characterize the I/O demands of data-intensive scientific applications. This paper analyzes several I/O traces collected from different distributed systems and concludes that correlations in parallel I/O inter-arrival times are inconsistent, either with little correlation or with evident and abundant correlations. Thus conventional Poisson or Markov arrival processes are inappropriate to model I/O arrivals in some applications. Instead, a new and generic model based on the μ-stable process is proposed and validated in this paper to accurately model parallel I/O burstiness in both workloads with little and strong correlations. This model can be used to generate reliable synthetic I/O sequences in simulation studies. Experimental results presented in this paper show that this model can capture the complex I/O behaviors of real storage systems more accurately and faithfully than conventional models, particularly for the burstiness characteristics in the parallel I/O workloads.