Assessing time coalescence techniques for the analysis of supercomputer logs

Abstract

This paper presents a novel approach to assess time coalescence techniques. These techniques are widely used to reconstruct the failure process of a system and to estimate dependability measurements from its event logs. The approach is based on the use of automatically generated logs, accompanied by the exact knowledge of the ground truth on the failure process. The assessment is conducted by comparing the presumed failure process, reconstructed via coalescence, with the ground truth. We focus on supercomputer logs, due to increasing importance of automatic event log analysis for these systems. Experimental results show how the approach allows to compare different time coalescence techniques and to identify their weaknesses with respect to given system settings. In addition, results revealed an interesting correlation between errors caused by the coalescence and errors in the estimation of dependability measurements.

Big Data Meets HPC Log Analytics: Scalable Approach to Understanding Systems at Extreme Scale

“…For the analysis of logs of large-scale systems, certain approaches apply filtering, followed by extraction and categorization of error events [14] [15] [16]. Other analyses use approaches such as time coalescing [17]. Some studies have focused on analysis of failure characteristics of specific subsystems or system components in HPC systems, such as disks [18], DRAM memory [18] [19] [20], graphical processing units (GPU) [21].…”

LogDiver

“…Prior research activities have centered on analyzing error logs [1][2][3][4][5][6] as well as some online analysis for patterns preceding a failure, and evaluated the accuracy and efficacy of anomaly detection and proactive response [12,13]. They have addressed one or more of the following issues: basic error characteristics [1,2,5], modeling and evaluation [6,14,15], failure prediction and proactive checkpointing [16,17]. There are many challenges in systematically studying large-scale systems using operational data, such as data availability, data collection/mining and fault/failure characterization.…”

“…Although today we understand the main characteristics of failures in supercomputing environments [1][2][3][4][5][6][7][8], the issue of job and application resiliency has been less well-studied. Modern supercomputers are equipped with fault-tolerant infrastructures that are capable of protecting job and application executions from failures due to either hardware or software problems.…”

Cost-Benefit Analysis of Virtualizing Batch Systems: Performance-Energy-Dependability Trade-Offs

“…The distributions of the transitions and the weights for the models are reported in Table I. They have been computed by means of workload analysis [14], log analysis [15], and linear regression [16]. Simulations are performed by means of SPNP [17] with the method of batch means, so achieving steady state behavior [18], with 95% confidence intervals and maximum half-width of 5%.…”