So far, everything works pretty well, I stopped extensively testing it and now the router runs in ‘every day work mode’ meaning there are 2-4 wireless and 1 wired client connected, none of them consumes a lot of bandwidth, just typical usage. I checked the log, the router did not restart for several days. The router runs switches periodically from/to repeater / wired mode, AdGuard uses 10 host files, the screenshot is attached.
This feature is the one I waited/wanted for the most.
I installed several additional packages such as the frond end for MWAN3 and made some changes in the config (which probably means that I slightly changed the original GL config).
I tried to tether a smartphone, it worked to some extended, meaning it started working fine and stop a bit later. To achieve that I had to modify TTL and hope limit counter for all interfaces.
But there is something that bothers me - when I capture the packets leaving the WAN, I see different TTL and this is very strange (although it has been reported by another user once).
The packets with incorrect TTL coming from MWAN3 which uses ping for check the state of specific interfaces. And the custom setting of TTL for POSTROUTING does not affect the TTL of the packets sent by MWAN3.
This is something I want to look into later, I also use GL-AR750S-Ext / Slate, running OpenWrt 18.06 with MWAN3 (and such) and in that case TTL is set correctly, the tethering works without any glitches.
This is probably all what I can say about AXT1800 right now.
I am not sure how I would compare Beryl with AXT1800. I would say that Beryl has a bit stronger signal than GL-AR750S, more memory, faster WiFi and a bit faster CPU. The best thing about Beryl is that it runs well OpenWrt while AXT1800 cannot (and this is important for me). The best thing about AXT180 that it has even more memory, faster CPU, faster WiFi and it can run AdGuard (and this is important for me). That is why I have both. But for you it is probably better just wait for a little longer before you can decide which router to get.
Plus, it would be useful to know what are your priorities in this regard:
a) run original OpenWrt;
b) run DNS filtering (AdGuard);
c) and so on.
Some results of speed test. It would be better to use a dedicated iperf server to test the speed but this time I used att speed test and fast.com. The router was connected (WAN) to pfsense , pfsense is connected to fiber / 1Gb line, the client (WIN10) was connected using LAN port of the router. All 3 devices run MWAN3 and some other services:
I was not satisfied with its performance while technically you can install it (may be I am a bit spoiled by pfsense running on mini PC (8GB RAM/ 128GB SSD), it has a similar plugin called pfBlockerNG).
I am also not sure, I have not used adblock. I have no doubts it would block the stuff, the question is how many entries can the block list include and how much it will tax the CPU (which in turn should affect the overall performance of the router).
I’m using the Stephen Black list, so about 66,000 sites. I’m using less than half the RAM and not noticing any ill effects. On my home router (also dual core and 256mb) I’m blocking about 200,000 sites without an issue, so I think it would be the same. Anyway, I think wireguard and wireless would be reasons to move up, but not necessarily ad blocking.
You can see the names of DNS blocklists and the numbers of the rules in each of them I use in AXT1800 on a screenshot shown in one of my previous posts.
It gives 402294 rules, without duplicates. I prefer one large list than many small ones, because as @elorimer said, you will have a lot of overlap unless they are designed to work together with other lists.
Thanks for the link, I see this is the list from Adblock Plus. Honestly speaking, I am mostly interested in blocking Windows stuff, AdGuard uses nextdns server as the upstream DNS server which does all DNS filtering heavy lifting.
Most likely, but it should not affect the performance (while the RAM consumption - yes).
I will try adblock on Beryl a bit later, right now I am flashing it with ROOTer (MT1300 is fully supported and that is the great news) which is much more suited for work with LTE modems (which I am pretty interested in).
It will affect both ram and performance. It still has to scan the filters, so if you have 10k filters duplicated over 4 lists, that is 40k filters it will process, instead of 10k if it was a consolidated list. In that example 4 times slower.
I don’t like 3rd party services, its another service i need to trust with my data. Nextdns has to process all your requests, they basically see everything you do on the net.
I seriously doubt. As soon as you found the entry in the first list there is no need to continue searching for this particular name since you already know that it has to blocked.
In addition to that If I would be a programmer who has to design a DNS filter I would join all lists into the one in the RAM (got rid of duplicates at this step), build a hash table and use some sort of quick binary search over the whole list.
True. It also true that ANY DNS server which receives your requests naturally sees them.
Yes, security comes with some cost. For example, you can point to nexdns your cell phone (if it allows private DNS, Samsung S10 allows) and all dns requests will be examined by nexdns.
If one looks at how much telemetry is sent/received by Android he could be really surprised, since in my view, OSs for mobile devices is the least audited/ user controlled kind of OSs.
)