Modeling high-throughput applications for in situ analytics

Abstract

With the goal of performing exascale computing, the importance of input/output (I/O) management becomes more and more critical to maintain system performance. While the computing capacities of machines are getting higher, the I/O capabilities of systems do not increase as fast. We are able to generate more data but unable to manage them efficiently due to variability of I/O performance. Limiting the requests to the parallel file system (PFS) becomes necessary. To address this issue, new strategies are being developed such as online in situ analysis. The idea is to overcome the limitations of basic postmortem data analysis where the data have to be stored on PFS first and processed later. There are several software solutions that allow users to specifically dedicate nodes for analysis of data and distribute the computation tasks over different sets of nodes. Thus far, they rely on a manual resource partitioning and allocation by the user of tasks (simulations, analysis). In this work, we propose a memory-constraint modelization for in situ analysis. We use this model to provide different scheduling policies to determine both the number of resources that should be dedicated to analysis functions and that schedule efficiently these functions. We evaluate them and show the importance of considering memory constraints in the model. Finally, we discuss the different challenges that have to be addressed to build automatic tools for in situ analytics.

SIM-SITU: A Framework for the Faithful Simulation of in situ Processing

“…Only a few recent attempts has leveraged simulation to enable the exploration of the in situ parameter space. Aupy et al designed a numerical simulator [10] that measures evaluation metrics for scheduling decisions by solving optimization problems on resource allocation and partitioning for an in situ analysis set. The simulator used a predetermined set of parameters to study the impact of the in situ analyses that are scheduled on the performance of the entire in situ execution.…”

“…This user can act on two parameters to execute the desired amount of analysis: the frequency and the cost of one execution of the analysis/visualization component. For instance, if the main simulation loop is executed 8,000 times and 400 units of analysis have to be performed, (at least) four (frequency, cost) configurations can be envisioned: (20,1), (200,10), (500, 25), and (1000, 50). The (500, 25) configuration means that 25 units of analysis work are performed every 500 iterations.…”

SIM-SITU: A Framework for the Faithful Simulation of in-situ Workflows

“…Different approaches have been proposed in the literature to ascertain the performance gains brought by an in-situ (or in-transit) execution of a given scientific workflow application and determine the best configuration deployment of its components on a given target platform. We distinguish these approaches depending on whether they rely on actual experiments [3][4][5][6][7] or resort to simulation [8][9][10] to evaluate the performance of in-situ workflows. The former is intrinsically time-and resource-consuming while the latter may suffer from simplification biases when the abstract versions of the in-situ workflow components are developed.…”

“…Only a few recent attempts has leveraged simulation to enable the exploration of the in-situ parameter space. Aupy et al designed a numerical simulator [9] that measures evaluation metrics for scheduling decisions by solving optimization problems on resource allocation and partitioning for an in-situ analysis set. The simulator used a predetermined set of simulation parameters to study the impact of the in-situ analyses that are scheduled on the performance of the entire in-situ execution.…”