Slowly but surely working on a #backlog of #tv , #films and #youtube #videos . It might literally be #christmas before I have fully caught up. On the plus side, my #cache #systems are working correctly!
Interesting #discussion on #cache #control in #http for #cors https://httptoolkit.tech/blog/cache-your.. Never even really thought about it.
Spent some time setting up #dnsmasq for my #local #desktop - turns out that it was #bypassing the #dns #cache with `no-resolv`: https://help.ubuntu.com/community/Dnsmas.. Seems to be worth it so far, should massively reduce #traffic !
This is a great #article about various #cache #issues and #bugs experienced at #twitter https://danluu.com/cache-incidents/ When #cache works, it is awesome. When it #fails , it can be #catastrophic to your #system or #service ... It's something that *needs* to be done 100% correctly!
@barray on Wed Sep 22 20:58:27 UTC 2021 said: &eLooks like there isn't a setting for #cache size in #deadsocial - so I will likely need to #code one from scratch: https://github.com/danielbarry/d3ad-soci.. I remember roughly in the code where that will need to happen though...I think I need to greatly #reduce the #deadsocial #cache size later, it's too large for the #available #ram I have on this #server ... It's currently using about 5MB out of the dedicated 6.4MB - this could be what's causing the occasional crashing during #builds !
I think I need to greatly #reduce the #deadsocial #cache size later, it's too large for the #available #ram I have on this #server ... It's currently using about 5MB out of the dedicated 6.4MB - this could be what's causing the occasional crashing during #builds !
Interesting, looks like #ibm 's approach to the #cache and #sharedmemory problem is just to do away with L3 and L4: https://www.anandtech.com/show/16924/did.. Will definitely be watching this space.
Awesome, the #wd #budget #nvme #ssd is getting cheaper parts swapped in that silently reduce #writespeeds - probably due to the #global #chipshortage https://arstechnica.com/gadgets/2021/08/.. Apparently once the #cache fills up the speed of the drive halves!
Nice discussion on whether #halfprecision #floats are worth the effort or not, specifically using the #gpu https://futhark-lang.org/blog/2021-08-05.. Turns out the #speedup really isn't so much. I suspect it *may* be better when you have large vectors, you can literally keep twice as much in #cache and larger models in #ram ... But still, it's worth checking whether you actually get the expected speed-up or not.
Personally had some experience with #cgi scripting, and all I can say is that it is slow as balls: https://landchad.net/cgi.html I've had much better experience with just using #nginx as a #reverseproxy to some #java #server running locally (exactly the configuration of this server). Currently the back-end can process a request in less than 1ms (ignoring other latencies). CGI on the other hand eats CPU and RAM for breakfast. The only saving grace is the #cache it can generate.