* * * * *
                                        
               How does TLS use less CPU than plain TCP, part II
                                        
I noticed over the past couple of days that the CPU (Central Processing Unit)
utilization were similar between the two services, but looking at the logs it
seemed my gopher server [1] git hit with several bots. So last night just as
the logs reset (a weekly job) I decided to restart the two services so as to
match CPU utilization and requests. Twelve hour later and … 

Table: CPU utilization after 12 hours
	CPU	requests
gopher	0:28	175
gemini	0:26	1744

A week and a half [2] after calling “Mission Complete” and apparently not.
Even worse, my gopher server is using a bit more CPU than Gemini server [3]
while getting one-tenth the number of requests. Something is off here.

I checked the requests and both receive over 90% requests for my blog, so I'm
confident that it's something to do with that. Both use the same request
framework [4], both use the same code to read blog entries, both use the same
code to parse the HTML (HyperText Markup Language), so it's not that. The
only difference left is the formatting of each entry. The gopher server does
wrap the text to 80 columns (not needed for the Gemini server), and there's
quite a bit more code to deal with fixing the links … are those the causes?
It's hard to determine what's causing the excessive CPU utilization.

I'm going to have to think about how to determine the issue.

Heck, I'm going to have to think if this is even worth determining. I mean,
the CPU utilization is better [5], but it's still puzzling.

[1] gopher://gopher.conman.org:70/1
[2] gopher://gopher.conman.org/0Phlog:2024/06/12.1
[3] gemini://gemini.conman.org/
[4] https://github.com/spc476/lua-conmanorg/blob/master/lua/nfl.lua
[5] gopher://gopher.conman.org/0Phlog:2024/05/27.1

Email author at sean@conman.org

.