Omnia/Mox successor

General discussion
1

Hi,

I was thinking little bit about Omnia and Mox successors in next future (2-4 years), because soon or later first Omnias will die.

Successor should be combination of these two with possibility to upgrade basic configuration with common components from notebooks (Turris team can give list with recommnded parameters), which will be in EOL in that time and possibility of using some of Mox modules (ecology + price savings).

So, the device/mainboard should be like:

CPU – some model with decent power

RAM – motherboard should be without RAM, there should be one one (or two) slot for notebook RAM modules (possibly DDR4)*.
*Optionally sold by Turris team for those who will not have some.

Storage – at least one M.2 slot (at least one NVMe)* + at least one SATA port for 2,5” HDD/SSD (with possibility of using adaptor for M.2 (SATA) / mSATA adaptor in this port)*.
*Optionally sold by Turris team for those who will not have some.

Ethernet – at least with 2,5Gbps support, but better with 10Gbps (WAN+4x LAN on motherboard at least) and PoE support (at least passive).

Wireless - some basic WiFi module on board (802.11ax in that time porbably – at least 4x4)
Place for internal USB(s) for Bluetooth, ZigBee, etc. - for modules for home automation…

Extension slots:
At least 2 SATA ports (on basic motherboard)
At least 2 (or better 3) mPCIe/M.2 slots for LTE, WiFi modules
Extension modules – connected via PCIe x16 cable (?) to ensure backward compatibility of some Mox modules.

Some modules example:
Storage module – with 4 other SATA connectors for HDD´s
Switch module – with additional 4-16 Ethernet ports
SFP module – with better support than today. Turris team should seriously cooperate with internet providers to ensure compatibility with their networks (e.g. Orange, T-Mobile, Vodafone, UPC).

Case
Something like Turris Omnia with “floors” – all additional modules should go down under main motherboard interconnected with cable
Orifices for 8 antennas (for full 8x8 802.11ax)
Extra buttons for: WiFi (switch on/off), AdBlock (switch on/off) and LEDs dimming (like on Omnia)

Power supply – none – only some transformer (to ensure correct supply voltage 12/19V?) with adaptor set for common notebook connectors (HP, Dell, Acer, Asus,…) - in basic package*
*Optionally basic power adaptor sold by Turris team for those who will not have some.

Other mainboard components- e.g. crypto chip, etc. which functionality was proven by time…

Software
OpenWRT like it is today with integrated AdBlock (with GUI suitable for non experienced user) out of the box.

What do you think, it is time to start another Indiegogo campaign? :innocent:

2

If nothing else, the plan sounds quite expensive. I’m no expert on this, but when I’d want to buy just a switch with two or more ports faster than 1G, I’m currently getting at least around the price of Omnia. The proposal contains much more than such a switch. Well, perhaps in farther future these features will get less expensive and more utilizable in real life.

1 Like
3

Well I would add at least 3 miniPCIE for

a) 1x WIFI 2 GHz 802.11abg or better wifi card just for legacy compatibility with home old devices
b) 1x WIFI 5 GHz 802.11ac for surrent defivices
c) 1x WIFI 6GHhz or Wave2 802.11ax

Reason for this is that once you connect legacy device to 802.11ax it get decreased to lower speed so really all 3 WIFI networks will be necessary for specified reasons.

So list of features is far to more than ordinary router so lets abandon ARM architecture altogether and get to amd64 / x86. most likely @michalko58 will end up with some x86 motherboard anyway. And yes it will be costly, but that power like AES-NI ! Altough I had not seen any board with 3x miniPCIE so far :frowning:

4

As much as I like your idea, this is going to cost crazy amount of money.

What I think would help most power users and is cost effective per unit is simply more raw horsepower. Even doubling those A53 cores would go a long way, but I’m pretty sure Cortex A72 is not out of the question (as evidenced by RPi 4).

Every modular part will make it more expensive as you cannot order in bulk.

5

Unless there are significant developments with SoC that would properly handle true 1Gbps+ throughput it would require at least

  • quadcore CPU

and with the current designs that would needs active cooling and subsequent a different form factor, power supply, board and so on

The matter came up earlier in this thread

There is nothing the TOS team can do as it is simply not in their scope of work

6

not sure, but if you want all these options, is’nt it better to build your own with a mini atx board? I mean, this company is way to small to have this produced in serious ammounts to keep the cost a bit in hand.

1 Like
7

Well this all is just an idea, but I thought that, if there won´t be any RAM, memory, etc., it should be cheaper. If the 10Gbps ethernet is still so expensive, it could be as optional module.

Maybe, if Turris team will cooperate with some large operator in their country (T-Mobile, O2, Vodafone or with all of them at once), their device can be offered in their networks (operators can offer it for rent) and everything can be cheaper.

Powerfull modular router (e.g with exchangeable modem in SFP cage or exchangeable modem module) with affordable price (100-200,- € for basic, but still powerfull, module), which could be used accross all operators and networks (fiber, vdsl, LTE,…), with software support is dream of all users. :wink:

And something like this should be supported by European Union, because lifetime sw support, exchangeable parts which ensure easy repairs (+ second chance for hw parts from old notebooks) and universal device for every single operator is not only ecologically friendly, but the price could be lower too. And as a bonus, there will be finally network device “made in EU” without backdoors from usa, china, russia (no offence to citizens of that countries).

As a single user I am not able to do this, but some company with Turris team can. :muscle:

8

Here I see mainly the problem with costs again. I believe most of common people get a very cheap router (here in CZ at least) and the internet seems to work just fine for them; it will be quite difficult to persuade them that something with way higher price will make a sufficient difference. Even if they don’t pay it directly, in the end it’s them who bears the cost. For common people it seems harder to benefit from a large part of the added value… though I could see hope in the security aspects due to updates. (But I don’t think updates themselves need expensive HW; that seems more about expenses on SW maintenance and testing, and risks of breakage.)

T-Mobile and O2/Cetin are even members of cz.nic, but I don’t expect that would by itself make a large difference given my reasoning above. They might theoretically decide to “force” it on their customers somehow, but what would be their motivation for the costs?

EU grants :thinking: there I might see more hope, at least theoretically. Well, I’m not even part of the Turris team, so I just make conjectures :slight_smile:

1 Like
9

Is it just a router or modem/router combo? If the latter an ISP would probably be more inclined to subside a combo device than just a router.

The incentive for an ISP to provide subsidised CPE is probably more in the cost mitigation for maintenance and/or customer support. Often those combo devices (modem/router) provide a management backchannel in the devices for the ISP to:

  • push firmware updates that been tested/approved
  • adjust modem/line parameters
  • monitor device and/or subscriber line performance/issues
  • remote debug the device if the subscriber faces issues

Which could even be a disadvantage and lead to compliance litigation with those vendors currently supplying CPE to the ISP, likely going to challenge if they were about to loose the deal with the ISP.

10

I am from Slovakia, so I know the situation with network device cost quite well. :slightly_smiling_face: Internet providers in Slovakia offer mostly devices for rent for price around 2,- Euro per month.

But video streaming in 4k etc. for more devices needs more powerful hardware.

We will seee what future brings to us.

2 Likes
11

For xDSL I typically see one device that does everything (including WiFi), though I can’t say I “have a large sample”.

From Turris experience the updates are a bit risky on maintenance and support, due to (the risk of) occasional bugs sneaking in. But supporting common users has the advantage that they typically leave almost every setting in default, so testing is much easier/surer.

12

Probably the same with g.fast and cable devices.

13

No. The new device should only have the following features:

  • 1*5GBE RJ45 with support for PoE+ for connection to network backhaul
  • 1*5GBE RJ45 for daisy-chaining access points
  • microSD-cardreader for OS
  • LEDs like Omnia’s with hardwarebutton and schnapps-integration (working fine on TO, is a pain on MOX with only one LED)
  • 2* M.2 interfaces instead of the 3 mPCIes @Twinkie mentioned (15/6GHz with WiFi6.1, 12,4GHz for Wi-Fi6) as soon as Wi-Fi 6/6.1 is supported by openwrt and M.2-cards are available. mPCIe is not fast enough for those cards (maybe for Wi-Fi 6 2,4GHz 4×4, but not for the others).
  • Metal case (for better heat dissipation than MOX) with 12 optional holes for antennas (8x8 + 4x4) with directly integrated wall mount - for Omnia you need to buy a separate mount, MOX doesn’t have one at all… :nauseated_face:

That is what is needed for a good access point.
If Omnia doesn’t fulfil your needs for a router/smarthome centre/NAS/server/…, skip searching for an embedded device. At least @Twinkie and me did and both of us ended building or own or buying a X86 device.
You can even include SFP+ for VDSL/g.fast or direct fibre access.

There are some APUs that have 3 mPCIe slots, have a look at the manufacturers website

14

I checked it very briefly but it looks like 3 minipcie is actually msata that has same connector but differen wires connected as referenced only 2 minipcie:

2 miniPCI express (one with SIM socket for 3G modem).
15

I went for this but so far I did not get to make ready to use yet as I am facing multiple problems, main problem is I was unable to compile succesfully custom kernel with LXC options enabled for 18.06.4 and even recent 19.07.0 only success with compiling master. I use debian buster to compile and maybe trying to use something obsolete will solve the problem. So far but master come with new version of kernel literally every week so unless taking kmods from local repo you are unable to install anything that needs aditional kmods. It is simply hell and there is not resolution for that as openwrt community simply says update is not possible as it always broke something. It is actually literally impossible. So you are forced to make new installation every time (installing and configuring all the ipk again and again) which is excessively time consuming.

DSC09425_resize DSC09665_resize DSC09667_resize

But from recent post it is obvious that more and more people that grown on TOS looking for more powerfulll and advanced device, My feeling is that there will be growing demand for such device in a future as internet and WIFI speeds advances what will be answer I don’t know. For geeks that does not mind pay the price for features they want it could be probably x86/amd64 board. But openwrt is god the pain. And lack of enough minipcie for all 3 wifi network types that I posted up is also another story.

2 Likes
16

I noticed that M.2 is necessary, but how about the kernel driver source support ? Is there already available for 802.11ax ? AFAIK Intel WIFI cards usually does not support Access Point mode so it is possible to use only on client side. My overall impression is even WIFI6 will comes sooner or later, there is not time for this, yet. But more and more devices will come to the market in coming days.

17

Did you try to run a decent distro with openwrt inside a vm?

You are right - sry, I remembered it incorrectly. How many mPCIe has your turris 1.x? I believe best is to split up the responsibilities in a network - 1 router that only does the routing or maybe also running some containers and a VPN server but not serving Wi-Fi and in addition to that several aps that only serve Wi-Fi. And maybe in addition to that a 10GBE backhaul :slight_smile:

18
19

Hope we will not be archived as off topic as this really is off topic, but anyway.
Well this chinese aluminium case device got Intel Braswell 4 core CPU with AES-NI and I put 240GB and split into 7-8 partition with Win10, Ubuntu, Debian Buster, Openwrt, data partition on BTRFS with snapper enabled and finally swap partition. So yes I running debian buster on it and also boot openwrt kernel from debian via grub2, so far I was unable to make boot openwrt from BTRFS so I kept on ext4. Don’t know what is the cause as I compiled BTRFS kmod into custom kernel together with LXC CGroups options. But I was unable to boot against openwrt root on BTRFS.

In my current Turris I have two minipcie with 2GHz and 5GHz 802.11ac card so for now I am covered but 802.11ax will come and more and more devices will come on the market soon.

My idea was to use this amd64 device as main router and transfer all the LXC’s with debian and pihole together with all other services like OpenVPN, shadowsocks, shellinabox to this powerful device with use of that powerfull Compex WLE card and then degrade my original Turris 1.0 to plain access point.

Sorry to say that as I really love my Turris router and have lot useful services running on it but as I already wrote in different topic, powerpcspe support was dropped from debian, it lacks quite a power for me now and even TOS 4 early testing is now available, final release is still in the fog for me. Togetger all of this means that it will became obsolete sooner or later as more and more technological innovations take place. But anyway I learned a LOT running it and I am thankful for that.

I have to laugh when people here compain with every release that this and that does not work. Those should try to compile your own openwrt and try make it running, it is huge I mean very huge task !

To configure regular buster as router and access point is also possible way but in that case I woulh have to put even more time into that and will lost LUCI and have to configure EVERYTHING myself which okay, I would learn a lot but not sure if got enought time to that to learn configure everything from command line.

20

That would be sensible but is currently not the concept for SOHO CPE whereas everything gets crammed into a single unit (consumer convenience).
What you are describing is more of a industrial grade (in terms of today’s perception) setup.

True 1+GbE connectivity is currently not achievable with SoC that are typically leveraged in SOHO CPE.
1+GbE connectivity might be a future necessity in a SOHO environment but then either it requires substantial development in the capabilities of SoC or else amd64 arch needs to be deployed and that presents a different challenge.