default_cleanup: Default functions to help determine which cluster node to use

In burchill/cs: Functions for working with Rochester's CS cluster

Default functions to help determine which cluster node to use

Description

These functions just apply certain default assumptions to the node information data frames generated from get_node_data:

Usage

default_cleanup(df)

default_filter(df, threshold = 0.7)

pick_n_best_nodes(df, n)

Arguments

df	a dataframe of node information, as produced by get_node_data
threshold	the threshold for the proportion of memory currently available by which to exclude nodes (e.g., 0.7 excludes nodes with less than 70% of their memory currently available)
n	the number of nodes to select

Details

• default_cleanup turns the important numeric values in the data frame (from get_node_data) from character columns to actual numeric columns, removing the "kb" from the columns pertaining to memory, and adding a column called percent_free which is the proportion of available memory out of the total memory.

• default_filter simply removes nodes that aren't "free," have numbers greater than 64, or have a proportion of free memory (percent_free) less than a certain threshold.

• pick_n_best_nodes returns n remaining nodes that have the most available memory.

burchill/cs documentation built on May 28, 2023, 1:29 p.m.