Member OP
Enough with the fake benchmarks! Sysbench is a lie and you know it!
Here is what actually matters! Compile something real! Run something that hits disk and network and CPU all together! Not some synthetic score that the host optimized for!
I propose we make a standard test! Everyone compiles the same project! Same flags! Same version! Then we post real numbers!
Who is with me!
Member
¡this is very good idea! ¿why we trust numbers that provider can manipulate?
Jajaja I try to write long explanation of compiler flags and cache effects but no, is too much, I delete it
Better: pick one project everyone knows. nginx? No, too optimized. Sqlite? Maybe. Something with autotools so we see configure time too, that hits disk random very hard
I use HostHatch and their benchmark is very fast but compile is slow, is very suspicious
Hej, good idea
On server I have malý server from RackNerd, compile time for nginx was 4 minutes 20 seconds
Into config I put standard flags, no optimization
No fancy benchmark, just time make -j2
Can we agree on project? I suggest curl, everyone uses, not too big, not too small
1. curl is decent choice but I propose better standard
A) specific version: curl 8.5.0
B) exact configure flags: ./configure --with-openssl --disable-shared
2. Measurement protocol
3. Wall clock time only, not CPU time
4. Must include download+extract in timing
5. Run twice, report second run (cache warm)
3. Reporting format
I. Provider name
Ii. Plan specs as advertised
Iii. Actual compile time
Iv. Any observed throttling
What flags did you use for the openssl build exactly
Curl is a solid choice for this test.
Same, I had slow compile on HostHatch too
