Using the parallel command to speed up log file processing

Summary

Processing large log files from the command line using tools such as cat, grep, and awk can be time consuming. By using the command parallel, processing time can be reduced significantly.

In this example, the time it took to search through 24 large log files went from 130 seconds to 26 second by using the parallel command:

[jemurraylinux-host 2020-06-01]$ time ls -1 smb_files.*.gz | parallel -P5 "zcat {} | grep scipy | wc" | awk '{sum+=$1} END {print sum}'
1557855

real	0m26.019s

Details

Example input data is 24 hours of SMB access logs:

[jemurraylinux-host 2020-06-01]$ ls -al smb_files.*.gz
-rw-r--r--. 1 root root 61153434 Jun  1 01:03 smb_files.00:00:00-01:00:00.log.gz
-rw-r--r--. 1 root root 73379126 Jun  1 02:03 smb_files.01:00:00-02:00:00.log.gz
-rw-r--r--. 1 root root 68187575 Jun  1 03:03 smb_files.02:00:00-03:00:00.log.gz
-rw-r--r--. 1 root root 64462449 Jun  1 04:03 smb_files.03:00:00-04:00:00.log.gz
-rw-r--r--. 1 root root 52274261 Jun  1 05:03 smb_files.04:00:00-05:00:00.log.gz
-rw-r--r--. 1 root root 49936464 Jun  1 06:03 smb_files.05:00:00-06:00:00.log.gz
-rw-r--r--. 1 root root 52163412 Jun  1 07:03 smb_files.06:00:00-07:00:00.log.gz
-rw-r--r--. 1 root root 62623665 Jun  1 08:03 smb_files.07:00:00-08:00:00.log.gz
-rw-r--r--. 1 root root 69905009 Jun  1 09:03 smb_files.08:00:00-09:00:00.log.gz
-rw-r--r--. 1 root root 80785420 Jun  1 10:04 smb_files.09:00:00-10:00:00.log.gz
-rw-r--r--. 1 root root 88264516 Jun  1 11:03 smb_files.10:00:00-11:00:00.log.gz
-rw-r--r--. 1 root root 84752054 Jun  1 12:03 smb_files.11:00:00-12:00:00.log.gz
-rw-r--r--. 1 root root 81681862 Jun  1 13:03 smb_files.12:00:00-13:00:00.log.gz
-rw-r--r--. 1 root root 78398116 Jun  1 14:03 smb_files.13:00:00-14:00:00.log.gz
-rw-r--r--. 1 root root 84952968 Jun  1 15:03 smb_files.14:00:00-15:00:00.log.gz
-rw-r--r--. 1 root root 89052806 Jun  1 16:04 smb_files.15:00:00-16:00:00.log.gz
-rw-r--r--. 1 root root 81828336 Jun  1 17:03 smb_files.16:00:00-17:00:00.log.gz
-rw-r--r--. 1 root root 69743081 Jun  1 18:03 smb_files.17:00:00-18:00:00.log.gz
-rw-r--r--. 1 root root 58067071 Jun  1 19:04 smb_files.18:00:00-19:00:00.log.gz
-rw-r--r--. 1 root root 50089323 Jun  1 20:03 smb_files.19:00:00-20:00:00.log.gz
-rw-r--r--. 1 root root 50349351 Jun  1 21:03 smb_files.20:00:00-21:00:00.log.gz
-rw-r--r--. 1 root root 48452189 Jun  1 22:03 smb_files.21:00:00-22:00:00.log.gz
-rw-r--r--. 1 root root 50196889 Jun  1 23:03 smb_files.22:00:00-23:00:00.log.gz
-rw-r--r--. 1 root root 49307426 Jun  2 00:03 smb_files.23:00:00-00:00:00.log.gz

These log files contain approximately 1.6GB of data:

[jemurraylinux-host 2020-06-01]$ ls -al smb_files.*.gz | awk '{sum+=$5} END {print sum}'
1600006803

Processing each file in serial takes about 2 minutes and 10 seconds:

[jemurraylinux-host 2020-06-01]$ time zcat smb_files.*.gz | grep scipy | wc
1557855 24925680 546752855

real	2m10.014s
user	2m25.545s
sys	0m12.419s

Using the parallel command reduces the processing time to 26 seconds:

[jemurraylinux-host 2020-06-01]$ time ls -1 smb_files.*.gz | parallel -P5 "zcat {} | grep scipy | wc" | awk '{sum+=$1} END {print sum}'
1557855

real	0m26.019s
user	2m17.196s
sys	0m12.157s

Be careful with the -P option on production systems. It specifies the number jobslots to run. 0 means as many as possible. Default is 100% which will run one job per CPU core on each machine. When processing a large number of files, it can easily fill up all available CPUs.


comments powered by Disqus