This seems to be because R2S boots from an SD card, no NVRAM type stuff.
The file system is split into 2 layers of system and configuration, to delete the configuration file layer then the network configuration is lost, otherwise everything is kept.
will moving to EX4 makes things better? or there is no difference in this regard?
The version of EXT4 does not have the function to restore factory settings.
But I'm not sure that upgrading the EXT4 firmware will preserve the current configuration, it has to do with how the system upgrade is implemented, I haven't used it so I don't have an answer.
so maybe this is a more generic question and not specific to the R2S, I think I'll ask in the configuration forum.
thanks anyway
I'd have to upload my build somewhere. As for the OS/Tool, I used etcher on openSUSE. Would be fine on Windows I'm sure.
++ @wevsty
If you went with the ext4 route, you can simply upgrade through the web UI or sysupgrade. In either of those methods, your /etc/config/* is retained, as well as any other files/paths defined in /etc/sysupgrade.conf.
I explain more about the sysupgrade.conf and rc.local files here: How to upgrade x86 OpenWrt? I have some examples there from my config, though in that thread I am referring to my setup on an x86_64 VM. Regardless, it’s also ext4 and the same principles apply with ext4 on the NanoPi R2S.
1 Like
Hi Guys!
Recently bought an R2S to serve as a home router for a 1000/50 FTTP connection which requires SQM/upload shaping. I tried OpenWRT and FriendlyWRT and seem to get better speeds with the latter, but cannot manage ~ >800mbps down with SQM enabled (using either fq_codel or cake) on either platform. Any general pointers or configuration tips would be excellent, and I can provide further info when I'm back home on my PC.
That would roughly match (respectively even surpass) the performance figures stated in the very first post of this thread:
SQM is (single-threaded and) rather CPU intensive, at 1 GBit/s WAN speed you're punching above your weight limit.
3 Likes
++ @Ziggy
I have tried to tell others the same through the various SQM tests (starting here) I posted earlier in this thread. Perhaps the thread is just getting too long now?
Either way, for everyone looking at this device for gigabit shaping, I just don’t see it ever being the right fit. There is only so much that software tuning can do before you ultimately just bottleneck at the CPU with SQM.
For gigabit shaping, you really need to be looking at x86_64 or possibly mevebu-based hardware. Haven’t used mvebu myself, but I can vouch for x86, as I use that now.
Bottom-line, there is more to shaping gigabit than just looking at interface port speed, saying “oooh, gigabit!” and hitting the purchase button. There are many devices that can route+NAT at gigabit speed because of efficient offloading. That means the CPU handles initial flow evaluation against firewall and route rules, but once evaluated subsequent traffic matching the same pattern can bypass the CPU (thus the offloading) and is why that is so fast.
SQM, as it stands today, requires what would otherwise be offloaded to instead be constantly evaluated by the CPU—it’s why SQM can do the magic it does. But this requires that your CPU be constantly ready to ramp up and handle those flows, especially when bursting up into the gigabit territory. To @slh’s point, this is CPU intensive and in my experience, ARM procs tend to struggle in that territory.
1 Like
After reading my prior post, if you’re still set on using the R2S, one option you have is to disable SQM on the ingress side since it really can’t handle your gigabit. But, you could enable it on the egress side to easily shape your 50mbps up.
Figured as much, thanks for the clarification!
Thanks for taking the time to reply. @_FailSafe I understand what you're saying. My Netgear R7000 was definitely not going to cut it, so I ended up buying the R2S prior to actually knowing I would need to shape my upload to avoid being throttled by the infrastructure's policer (which doesn't seem to happen on any other plans, just this new gigabit plan). Without shaping, I easily hit 950mbps+ with the R2S, on both OpenWRT/FriendlyWRT. I just wanted to double check that there were no 'tweaks' or something that I didn't know about, so I would know for sure that I wasn't leaving any performance on the table prior to moving to an x86/x64 system. I did try shaping egress only but it still seemed to cause an overall performance drop versus having SQM disabled.
EDIT: I really appreciate all the hard work @jayanta525 and everyone else on this forum is doing for this device!
1 Like
I want to encourage caution with a statement like that. Others will come here, read just that, and then wonder why it’s not shaping their 600mbps connection (arbitrary speed pick) even though that’s <1gbit. For instance, if you look back at some of my prior posts I linked to, you’ll clearly see I was having trouble shaping even my 480/24mbps docsis connection.
I would say this is likely a better device for shaping 400mbps ingress or less, though there might be some gray area between 350-450mbps based on fq_codel vs cake (and other settings therein).
Fair, fixed.
pls add a link to the 2-9 version
Has anyone ever tried this firmware ?
Looks like that is using the friendlyarm code mixed with openwrt-lean and other githubs. Also, seems very recent and it benchmarks 40mbps higher than Jayanta https://github.com/jayanta525/openwrt-nanopi-r2s
I'd like to get Jayanta's changes mainlined before using more friendlyarm code I think.
Hello, @andiohn This thread was created when OpenWrt was initially supported on the board.
The repo is based on friendlywrt kernel, but with mainline OpenWrt. I tried to keep the kernel sources as close to the vendor's kernel to provide maximum compatibility such that users can use this instead of FriendlyWrt which IMO was bloated with unnecessary packages.
As this board is now merged into mainline OpenWrt, the repo is no longer maintained.
1 Like
Awesome! I've been trying to find the images or how to add it. Can you help out at all? I'll type up what I learn and see if I can help out someone like me in the future.