RESUMEN
Hadoop MapReduce reactively detects and recovers faults after they occur based on the static heartbeat detection and the re-execution from scratch techniques. However, these techniques lead to excessive response time penalties and inefficient resource consumption during detection and recovery. Existing fault-tolerance solutions intend to mitigate the limitations without considering critical conditions such as fail-slow faults, the impact of faults at various infrastructure levels and the relationship between the detection and recovery stages. This paper analyses the response time under two main conditions: fail-stop and fail-slow, when they manifest with node, service, and the task at runtime. In addition, we focus on the relationship between the time for detecting and recovering faults. The experimental analysis is conducted on a real Hadoop cluster comprising MapReduce, YARN and HDFS frameworks. Our analysis shows that the recovery of a single fault leads to an average of 67.6% response time penalty. Even though the detection and recovery times are well-turned, data locality and resource availability must also be considered to obtain the optimum tolerance time and the lowest penalties.
RESUMEN
BACKGROUND: In task-state functional magnetic resonance imaging (fMRI), hemodynamic response (HDR) shapes help identify cognitive process(es) supported by a brain network. However, when distinguishable networks have similar time courses, the low temporal resolution of the HDRs may result in spatial and temporal blurring of these networks. The present study demonstrated how task-merging and multivariate analysis allows data-driven separation of working memory (WM) processes. This was achieved by combining a WM task with the Thought Generation Task (TGT), a task which also requires attention to internal representations but no overt behavioral response. METHODS: 69 adults completed one of two tasks: (1) a Sternberg WM task, whereby participants had to remember a string of letters over a 4-sec delay or no delay, and (2) the TGT task, whereby participants internally generated or listened to a function of an object. WM data were analyzed in isolation and then with the TGT data, using multi-experiment constrained principal component analysis for fMRI (fMRI-CPCA). The function of each network was interpreted by evaluating HDR shapes across conditions (within and between tasks). RESULTS: The multi-experiment analysis produced three WM networks involving frontoparietal connectivity; two of these were combined when the WM task was analyzed alone. Notably, one network exhibited HDRs consistent with volitional attention to internal representations in both tasks (i.e., strongest in WM trials with a maintenance phase and in TGT trials involving silent thought). This network was separated from visual attention and motor response networks in the multi-experiment analysis only. CONCLUSIONS: Task-merging and multivariate analysis allowed us to differentiate WM networks possibly underlying internal attention (maintenance), visual attention (encoding), and response processes. Further, it allowed postulation of the cognitive operations subserved by each network by providing HDR shapes. This approach facilitates characterization of network functions by allowing direct comparisons of activity across different cognitive domains.