How to Modify CPU Core Usage in lcore for Large Instances

WHAT TO EXPECT

There is a known issue where users with large instance types experience reduced performance. To solve this issue, users can manually modify their CPU core usage to ensure an optimal thread configuration.

Prerequisites

• Aim to use only one CPU per core to maximize efficiency.

• Restrict usage to CPUs from the same socket, most likely from SOCKET 0.

• Try not to use more than 14 CPUs as listed in the lcores setting.

If you have any questions or concerns, please contact support@swxtch.io before attempting reconfiguration.

1. Check your core topology by using the following command on your cloudSwXtch VM:

   lscpu --extend --all

   Example Output:

   CPU NODE SOCKET CORE L1d:L1i:L2:L3 ONLINE    MAXMHZ   MINMHZ
     0    0      0    0 0:0:0:0          yes 2800.0000 800.0000
     1    0      0    0 0:0:0:0          yes 2800.0000 800.0000
     2    0      0    1 1:1:1:0          yes 2800.0000 800.0000
     3    0      0    1 1:1:1:0          yes 2800.0000 800.0000
     4    0      0    2 2:2:2:0          yes 2800.0000 800.0000
     5    0      0    2 2:2:2:0          yes 2800.0000 800.0000
     6    0      0    3 3:3:3:0          yes 2800.0000 800.0000
     7    0      0    3 3:3:3:0          yes 2800.0000 800.0000

2. Adjust your thread assignment by editing the /swxtch/lcores.json configuration file. This file specifies which CPU your application should use.

   This is an example of the format you can use in lcores.json:

   {
       "lcores": "1,3,4,6,8-12"
   }