Part of me thinks this is a fantastic idea, subject to some of the points already made (multiple LAN ports, 10G capable on WAN, SFP cage option, sub $100 etc etc).
But a small part of me also thinks that this could be a mistake. There are so many conflicting requirements that have already been suggested that could easily shove the cost way up beyond the suggested price point.
Personally I'm a bit fearful that if such a device ends up being a lame duck, it could hurt the reputation of OpenWRT quite badly.
port selection is good (yes, i know, many will disagree)
no mater how many ports you put in design, there will always be: i need one more
so,
1G for WAN
2.5G for managed switch is very good decision
this way, even multiple VLANs on one downstream port will result in 1G most of time
which is balanced with WAN speed
and the inter VLAN routing will not suffer, again, because 2.5G could handle 2x1G on managed switch
nice design for advanced home/small office network
As shipped by Zyxel, the XGS1010-12 it isn't managed, but the hardware is almost identical to the managed version of the same switch (XGS1210-12), and both can run OpenWRT (edit: There now appears to also be a XGS1010-12 V2 (XGS1010-12-ZZ0102F) available so caveat emptor...).
But there are no guarantees of course And I believe the XGS1010-12 comes without a reset button and mac addresses, since those are unnecessary on an unmanaged switch.
FWIW I also think that a single OpenWrt designed router+AP would be coming a little short with only 2 ethernet ports. The majority of users would have to buy a managed switch and a decent one would cost them half as much as the router or more, greatly diminishing the attractiveness of a sub 100 USD device, plus they'd have an extra box to take care of.
All-in-one router+AP+switch are so popular because they're the most affordable and practical solution for the majority of people. Filogic based all-in-one's will be plentiful and go down in price as time passes.
I don't say that for my own benefit as I own a few managed switches. I'd just like that this product will be as successful as possible.
Probably the best solution would be to offer 2 devices, one with 2 ports and the other with 5, but of course that would add to the complexity of the project. You could start offering the 2 ports one first and later add the one with 5 ports.
Great idea, give me two! (at least!) If this gets far enough along, I'd be willing to buy in advance, if crowdfunding can help the effort. I would second the idea of adding a switch, but I'd be in either way.
First I would like to start by wishing Happy Birthday to OpenWRT. It's my absolute favorite to run my networking for years now! I have it running for everyone I have set it for by picking and choosing components of the setup I mentioned below all running OpenWRT.
Router specifications feedback
However this router/AP is not something I would buy.
It's simply not desirable except for a very small group of people
As a router: worse than my workhorse R7800 :
Plus:
2.5GbE - Useless as the other one is 1GbE; If there is a use-case for having it a single 2.5GbE (I assume only as LAN) and I don't know about it (while knowing a more than the average person about networking) think about what that average person will see when looking at the specs
m.2 - Maybe for a WiFi network card to add 6E;
used for storage no for me, no for average person, required for some advanced users:
But why do you even need storage on a router (more on that later... seriously why?)
Size is 2042 for NVMe. I don't want to buy yet another format. I want all my NVMe to be 2280 so I can move them around as I please.
Minus:
3x1GbE ports
(2x2 2.4 GHz + 3x3 5Ghz) vs (4x4:4 + 4x4:4*ish) on R7800 -
As an AP: worse than my Cudy WR2100 which I got a whole bunch of them for $30 a piece when they were still around
Plus:
2.5GbE, again useless, you can't realistically push between router and clients more than 1GbE over WiFi 5 anyway as there will be a lot of signal loss on the WiFi anyway.
(2x2 2.4 GHz + 3x3 5Ghz) vs (4x4:4 + 4x4:4) on WR2100
Current ideal setup
Below is my ideal setup for 1GbE + WiFi 5 from which I pick and choose components based on the needs of whoever I am doing the installation for:
Requirement: Expand WiFi coverage with Ethernet back-haul, set each AP on non overlapping channels
Why: Extenders and Mesh just add interference as they use WiFi for back-haul use only if your can't realistically run cables
Server
Requirement: Raid 1 storage [absolutely mandatory as I mentioned here], x86 just because it's the most flexible to find components runs absolutely anything in terms of software
Why: As discussed here or here and here, you should really, really get a separate server
Upgrades
Now let's talk real life upgrades, here is what I see as upgrades:
The router processing power so it can route the full 1GbE WAN
'peak' speed for one user using the entire channel at distance changes very little (around 11% improvement over 802.11ac
Keep all of the marketing hype in perspective:In order to take advantage of Wi-Fi 6 improvements, you need client devices that support Wi-Fi 6 and you must be very close to the router. Until this happens, Wi-Fi 5 will do just fine in most homes. Most of the speed advances in 802.11ax (MU-OFDMA) will NOT materialize until ALL client devices are 802.11ax
When Wi-Fi 6 CAN deliver the goods: If you have a brand new Wi-Fi 6 client device and a brand new Wi-Fi 6 router supporting DFS channels, and are using both in the same room (so both devices are very close to each other) there is a high likelihood that the two devices will negotiate an initial 160 MHz channel width (Windows laptops and Android; not iOS). Throughput can be as high as 80% of the 2401 Mbps PHY speed (or around 1900 Mbps) -- which is very nice! However, this only happens when the client device and router are very close to each other (in my testing, four feet away) and only if 1024-QAM can actually be used -- and once you start adding distance or walls, the two Wi-Fi 6 devices will 'slow down' significantly (use lower QAM levels) and communicate with each other at much closer to Wi-Fi 5 speeds.
As a continuation of my previous post here is how I see what will most common happen in the real world.
Let's start with the network architecture which will be in most home/small office deployments
Router
WAN <- ISP
WAN/LAN <- ISP OR Home Server
LAN <- Work Station
LAN <- Access Point
LAN <- Switch with the following clients
Access Point
Cameras
TV
Printer
etc.
WiFi
Now think about each of these components, what would their requirements be?
Then you can pick your product to target each of these product slots
Also think about that almost all the companies have multiple products to cover each of the slots
If you want to create a niche product, think what's the size of that niche?
Also notice below that I do not mention m.2 for storage. Why? Because you should really use a separate server running RAID 1 for storage. Either mount the storage from the server, or if your data is that disposable use an USB drive. One day your drive will fail, no matter which one you use.
On a long enough timeline the survival rate for everyone drops to zero
If having expansion slots is a limitation of the number of PCIe lanes, then maybe there is a way of having shared lanes. Like on a desktop motherboard where if you add an PCIe card it will disable the m.2 slot sharing those lanes.
Router
Top Tier
CPU fast enough to handle the routing below with SQM or 1-3 very demanding package(s)
Decent RAM to run a few additional packages
Some internal storage to store the packages
2 x SFP+ 10GbE
4(5) x 2.5 GbE (1/2 WAN + 2/3/4 LAN)
WiFi 6E / 7 with 4T4R for each band (preferably MediaTek) OR m.2 / miniPCIe with enough bandwidth to handle a future expansion card for WiFi 7
Sample use case:
Router
2.5 WAN <- ISP
2.5 WAN/LAN <- ISP / Access Point WiFi 6E / 7
2.5 LAN <- Access Point WiFi 6E / 7
2.5 LAN <- Access Point WiFi 6E / 7
10 GbE SFP <- Home Server via SFP copper
10 GbE SFP <- Switch via SFP copper with the following clients:
10 GbE SFP - Work Station via Fiber
1 GbE - Cameras
1 GbE - TV
Printer
etc.
WiFi 6E / 7
Second Tier
CPU fast enough to handle the routing below with SQM or 1-3 demanding package(s)
Decent RAM to run a few additional packages
Some internal storage to store the packages
1 x SFP+ 10GbE
4(5) x 2.5 GbE (1/2 WAN + 2/3/4 LAN)
WiFi 6E / 7 with 4T4R for each band (preferably MediaTek) OR m.2 / miniPCIe with enough bandwidth to handle a future expansion card for WiFi 7
Sample use case:
Router
2.5 WAN <- ISP
2.5 WAN/LAN <- ISP / Home Server
2.5 LAN <- Work Station
2.5 LAN <- Access Point WiFi 6E / 7
2.5 LAN <- Access Point WiFi 6E / 7
10 GbE SFP <- Switch via SFP with the following clients:
Cameras
TV
Printer
etc.
WiFi 6E / 7
Third Tier
CPU fast enough to handle the routing below with SQM or 1-3 average demand package(s)
Decent RAM to run a few additional packages
Some internal storage to store the packages
4(5) x 2.5 GbE (1/2 WAN + 3/4 LAN)
WiFi 6E / 7 with 4T4R for each band (preferably MediaTek) OR m.2 / miniPCIe with enough bandwidth to handle a future expansion card for WiFi 7
Sample use case: same as above
Third Tier Minimum
CPU fast enough to handle the routing below with SQM or 1-3 average demand packages
On the non LTS version has a miniPCIe slot on top, available bandwidth TBD
The 2 x m.2 slots (one for WiFi card) are on the bottom close to the case, they may or may not be usable to add an AP WiFi card
Fourth Tier
CPU fast enough to handle the routing below with SQM or some additional packages
Decent RAM to run a few additional packages
Some internal storage to store the packages and some logs
4(5) x 1GbE LAN
WiFi 5 / 6 with 4T4R for each band (preferably MediaTek)
Sample use case:
1 GbE WAN <- ISP
1 GbE WAN/LAN <- ISP / Home Server
1 GbE LAN <- Work Station
1 GbE LAN <- Access Point WiFi 5/6
1 GbE LAN <- Switch
Cameras
TV
Printer
etc.
Access Point
First Tier
CPU with hardware cryptography
Minimal RAM to run the AP
Minimal storage for a few packages
1x2.5 GbE with PoE
Can be powered from PoE and outlet
WiFi 6E / 7 with 4T4R for each band (preferably MediaTek) OR m.2 / miniPCIe with enough bandwidth to handle a future expansion card for WiFi 7
Second Tier
CPU with hardware cryptography
Minimal RAM to run the AP
Minimal storage for a few packages
1x1 GbE with PoE
Can be powered from PoE and outlet
WiFi 5 / 6 with 4T4R for each band (preferably MediaTek)
Third Tier
CPU with hardware cryptography
Minimal RAM to run the AP
Minimal storage for a few packages
1x1 GbE with PoE
Can be powered from PoE and outlet
WiFi 5 / 6 with 2T2R for each band (preferably MediaTek)
Edits:
Mentioning that the AP only needs to be able to powered using PoE and outlet (USBC is nice, but can be anything except directly plugged in the outlet), so you can use either one depending on the situation in the field
Removed the storage of logs from requirements, as users pointed out they are stored in RAM.
Reduced the requirements from (SQM and CPU/RAM intensive packages) to (SQM OR CPU/RAM intensive packages) because: If (as a user below metioned) using an 4+4 core RK3588 SQM tops out at about 1400+Mbps, it's only good enough for a 1GbE ISP. I do hope this only because it's so new and it lacks software optimizations. So:
the most possible top setup for MT7981B should be like:
SoC: MediaTek MT7981B
RAM: 1024MiB
Flash: SPI-NAND 128 MiB or 256 MiB
Ports:
- 5 x Gigabit Ethernet ports
- 2 x 2.5 Gigabit Ethernet ports
USB:
- 2 x M.2 slots for 5G modems via USB 3.0 hub, external USB 3.0 port
USB Port:
- 1 x USB 2.0
WiFi: MediaTek MT7976CN
Buttons: Reset, Mesh
LEDs: 2.4G, 5.8G, 4G/5G x 2, 4G/5G signal x 2, port x 7, sys, pwr
So I am not publishing my own "wish-list" as I consider this to be a game of trade-offs, where those that do are the ones that also do to get to make the choices; I also consider the resulting proposal pretty decent/attractive.
@NPeca75 offered a quite decent and IMHO convincing rationale for that, that helps if users want to add a switch.
Easy, a router is typically running 24/7, which is also desirable for a file server, so making the router also act as file server can conserve energy. And say if this is not used to share privacy sensitive files, but e.g. a music/video library where the risk of access by hostile parties is rather low this seems like a fine option. Also to implement persistent logging non-build-in/replaceable storage is desirable.
Well, for logs you really need to have replaceble storage, eMMC/NOR/NAND do not tolerate all that many write cycles and hence are best not used for cyclic stores like logs, IMHO.
For me, extra storage only for packages, when a router is powerful enough I might want to run some network related containers, like PiHole, Grafana, they really need some storage, with approach like ram log (just write maybe once a day) the life of storage can be long.
Happy birthday OpenWrt! But I don't like the proposed specs. I would really like to see the arguments why that design was chosen. In my opinion, it won't sell because it's not a good fit for any of the common purposes:
As a portable router: no SIM slot, no slot for WWAN card, CPU too powerful and power hungry, too much memory, too many debug features, lots of other unused features that use extra power, can't power it from a power bank or an ordinary USB phone charger (5V). Probably too big too. Lots of other (arguably better) options on the market.
As a router: All of the above, no switch, wifi soon outdated (if not already) and can't be replaced.
As a NAS: NVME storage too fast for it's CPU and networking, also too expensive at any size and difficult to plug-in/remove/swap, no USB3, no eSATA. Even if it had USB3 or eSATA, memory too small for zfs deduplication or any significant caching.
As a VPN-to-LAN point of access: CPU too weak, no mention of hardware crypto. Memory barely enough for a content-filtering proxy or TOR node.
As I see it, it's only good as a training device for embedded developers, but even then, 128 MB flash is not enough for dual-partition booting and AFAIK OpenWrt can't dual-boot kernel from one device (flash) and root from another (NVME), which only leaves NVME as the only possible boot device - something that most routers don't have and any low-level development (kernel, failsafe mode, blockmount) on this device won't directly benefit other platforms.
There are lots and lots of devices out there that this proposal is struggling to compete with and will soon be e-waste. Too bad nobody thinks of the future. What OpenWrt should focus on is a device that can be updated when FOSS is abandoned:
Remember 8/32 devices becoming unsupported? Have memory as SO-DIMM modules like old laptops. Have storage on SD card, like Raspberry Pi.
Remember mwlwifi driver becoming unmaintained? Have wifi as mPCIe removable cards.
