To use the title: Is it just me or Raspberry 4 (4GB RAM) considerably enhances Pihole's performances?
To be honest, I didn't notice any higher hardware resources consuption, according to top and Pihole's stats. But...
I was experiencing multiples performances issues previously, as explainded and described in my previous post. And your help was amazing to help me solve them!
But I have to admit, with the same exact config (just transferring the SD card from one Rapsi to the other) that my network is WAY FASTER (not just placebo effect) since I've bought a Raspi4 (was on Raspi3B+ before).
I noticed too that my custom-255,000-adlist isn't a problem anymore. On my Rapsi3B+, I was forced to use the default list (good enough but not sufficient to me) in order to get decent performances.
Does anybody could explain that? I'm only curious. May be the separated bus for Ethernet (I do notice too that upgrade from Rapsi repos is faster, in terms of MB/s)?
It is just you. The small amount of bandwidth required for DNS traffic is much less than that provided by anything as old as a Pi 1. Only the DNS traffic goes through Pi-Hole; it sees none of the data traffic.
For Pi-Hole purposes, a Pi-3B+ is complete overkill, and a Pi-4 is more than that.
It is just you. The small amount of bandwidth required for DNS traffic is much less than that provided by anything as old as a Pi 1. Only the DNS traffic goes through Pi-Hole; it sees none of the data traffic.
Was my thinking. Thanks.
But in fact, it's true and mesureable. Complex DNS request implicating CNAME resolution take much lesser time. But I, technically, can't explain it at time (IPv6 disable network wide, etc.).
Does processor or RAM amount can have any side-effect?
I hate people who at every update say it's much faster. But for once, it's my turn. xD
The extra RAM helps a lot. Having more space for the large lists without needing to swap to the SD card for all the processes involved is a bonus. Just running stock though shouldn't need more than the 1G RAM. I run on 512M on all stock and don't really notice it.
With no numbers to compare, its hard to say.
Should setup a test lab with the different models.
Me too am still running stock lists on one of the very first Pi's without any issues:
pi@noads:~ $ cat /proc/cpuinfo
processor : 0
model name : ARMv6-compatible processor rev 7 (v6l)
BogoMIPS : 697.95
Features : half thumb fastmult vfp edsp java tls
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x0
CPU part : 0xb76
CPU revision : 7
Hardware : BCM2835
Revision : 0003
Serial : 00000000dd68xxxx
And still have some 130MB free for some more lists:
pi@noads:~ $ free -h
total used free shared buff/cache available
Mem: 179M 41M 30M 7.2M 107M 83M
Swap: 99M 30M 69M
Yes, I think this is the case. I have two Pi-Hole pairs running (each with a 3B+ wired and Zero W wireless), each running unbound as a local resolver. I cannot tell any difference in the DNS performance between either of these platforms. The 3B+ has 1GB RAM, the Zero W has 512 MB and the processor on the 3B+ is much more powerful.
If there are speed differences with Pi-Hole, they are measured in msecs, and not noticeable. Where the more powerful SBC really makes a difference is in day to day management - system updates, debug logs, rebuilding gravity, terminal operations, etc. are all significantly faster on the 3B+ than on the Zero W.
I see a very slight speed difference loading the dashboard on the web GUI with the slower SBC - it is perhaps a half second slower on initial load.
