Addendum, minor corrections:
Addendum, minor corrections:
working hard apparently so many videos, ill report back with notes once i get some free time
A bit late, this video was made some time ago. Adding it for completeness.
It’s been a while since I made a video, Life has been busy but I would like to continue. Outside of a Looking Glass tutorial, what else would people like me to cover/do?
In the meantime I have started streaming some content on Twitch, it will mostly be gaming sessions with my mates, some programming stuff from time to time, perhaps even some server management when security permits.
A topic I’m interested in is proper benchmarking/testing of network and storage setups, along with the reasoning behind why a method of testing is valid. It’s actually a pretty complex topic with many gotcha’s depending of the use cases and technology that someone is trying to test against. It’s a big enough topic I haven’t had time to really sit down and make sense of any of it.
An example of some common tripups I’ve uncovered
-DD is single threaded. FIO is multithreaded and generally preferred as a more realistic test.
-The different types of random number generators have different limits to how many numbers they can actually produce per second.
-Systems with a cache or simple compression (ZFS) will give you test of your ram instead of your intended hardware if you’re not careful. I’m looking at you “if=/dev/zero”
-What kind of workload do you need to optimize for? A system optimized to read and write huge files as fast as possible is different from one where a ton of tiny changes are trying to happen all at once. Sequential writes or reads are pretty much never an issue for anyone regardless of optimization, but 4K random IO will separate the men from the boys.
I hardly expect a comprehensive treatise, but any practical insights you may have gathered from your experience would be interesting. At this point all I really know about the subject is a little bit about what I don’t know, though frankly it appears even highly experienced people are conflicted about the best way to go about things.
Another possible topic to explore would be simple things you’ve done to improve network hardware over the defaults, such as enabling hugepages. There may be some specially considerations with VM’s and network configuration that I’m completely ignorant of.
You seen to have a good knowledge of PCIe. I’d very much enjoy a series on the technical details of PCIe. From the electrical signalling over the transaction protocol up to OS access. But that might be only me. ;3
My understanding of PCIe is quite limited at this point, which is why I have been handing off my findings to those who are, such as Alex of RedHat. I do have a good understanding of digital electronics and embedded microcontrollers which helps when reading and understanding the datasheets, etc.