Wise words. Similarly, publicly accessible services shouldn't be supplied on a device dedicated to security.
Too bad the Raspberry Pi series have such a bottleneck with Ethernet
1 Like
There is a whole school of thought that says that WiFi AP should be decoupled from the router and from the access modem. All three of these have different standards evolution lifecycles. However, from a cost perspective, the low-cost need that has been driving the edge has not allowed for such a decomposition. You see some of that today probably with mesh WiFi and a little bit with DOCSIS 3.1 modems but mainly you have co-existence. To me that creates lots of issues. There could be some modular approach where CPU/RAM/flash could be shared with certain components being upgraded. But not sure if it's realistic to think about any such complexity sitting at the edge.
Yes, though there are plenty of ARM boards pushing into this space, like espressobin and ODROID boards and things, the real problem is they don't have the software development backing behind them that RPi has. I expect maybe the next RPi will have gig ethernet, it might hang off a USB3.0 but even still, it won't be a bottleneck anymore for this kind of application. An RPi with gigE (and removable flash!) and an 8 port smart switch like the one from TP-Link would make a very reasonable home router system.
https://www.amazon.com/TP-Link-Ethernet-Sheilded-Replacement-TL-SG108E/dp/B00K4DS5KU
The current version of that switch is pretty good, it had some issues in the older revisions.
1 Like
A Pi reference on "Pi Day" (3/14)
I see what you did there...
1 Like
After having spent time on Alibaba, per one post's suggestion, I quickly realized the lower quality of engineering there. 6 antennas on a router, instead of dual band antennas, boards that might not have been well designed, tested at scale, etc.
I continue to think that if we were to place a group order here and keep a tab open at an OEM, we could - as a large customer - influence the availability of an open architecture.
We could, in essence, create a home router, OpenWRT-ready version of the RPi.
This would use ath10k or mt76, would have just the right amount of RAM/flash required and in my view would not exceed the $50 price point. It would be cheaper than a loaded RPi with much higher quality of certain networking components (you don't need some ports there like HDMI, SD cards, audio outputs, etc.)
Already been done...
Yep, I run PCEngines 2C4 for my router and another for my firewalls, various services on a mix of Raspberry Pi of all generations, ODROID, and FreeBSD jails.
Coming back to the OPs original question, it is the second question that should be asked. I still haven't heard
What is the business plan?
Unless there is a compelling need in a sufficiently large market, no "recognized" manufacturer is going to touch this with a 10-foot pole. "The forest" could care less -- they're going to buy whatever their service provider offers them, or Amazon or Walmart or the like has cheap. The enthusiast community isn't going to buy a low-end router. Shoot, I can pick up 5-10 Archer C7 units in great condition off Craigslist today for $30 a pop around here. They aren't going to push more than 400 Mbps, but they're going to run circles around anything an ODM can put together for a price that competes, without even considering marketing, sales, and support costs.
2 Likes
That was my point in saying something at the high end of ARM or the low end of x86 is really more the issue. Pretty much everyone on this message board system even the new guys are doing something that would be considered "sophisticated" compared to what you typically can do with say a TP-LINK OEM firmware. Even if it's just setting up SQM or an OpenVPN connection to a provider.
The things that are "wrong" with consumer routers are:
- Easily Brickable, and requires soldering for serial
- Either too low powered / small storage, or too expensive and focusing on high end WiFi hardware rather than CPU
- Some kind of manufacturer lock-out so you can't install LEDE
- Undocumented bootloaders / flash layout and etc etc
You solve those problems with a low end x86, or with an espressobin type board and some manufacturer support to get the drivers going well.
I just don't see a purpose to trying to get into the $50 wifi router market, that's a market that already ships MILLIONS OF UNITS A MONTH many of which already work fine on LEDE like the C7
Even if you look at Meraki, I trust you the hardware cost may not even be more than $80-$100. The value is in the software and that's where the OpenWRT value is as well. But unless you standardize the hardware and do not have to spend cycles chasing down and reverse engineering locked drivers, new chipsets, RAM/CPU selections and going from device to device, this level of effort could be focused on making the software even better. Having established that kind of structure, the creativity of whoever is the end user, a hobbyist, an ISP, a small business that runs several APs, a system integrator can be unleashed on top. This, in my mind, will have a profound impact on helping networks become more software defined, more open source code, adopt NFV, etc.
I hate to burst your bubble, but hardware is moving in the other direction. Take a look at http://www.ti.com/product/cc3220 for example. The whole 802.11 and BLE stacks are in hardware, with hardware encryption. Unit cost is $5.71 in 2,500 quantities. Linux has a limited lifespan as there won't be a need for it soon.
To a router/AP manufacturer, the only value in software is keeping the support calls to a minimum.
This is IoT related where low power consumption and small size are key.
That's a different discussion.
No, it's a question of cost -- You consistently ignore commercial reality. You've never once given a business plan, or even identified your target market. Until you do that, this is a pointless discussion.
There are two dimensions here.
The one is that, in my humble view, it would be great if the OpenWRT/LEDE community could define its own ideal hardware reference architecture and concentrate a lot of its effort on that. That would save a ton of time chasing down a number of issues related to supporting a growing number of devices. At the same time, you read repeatedly on the forums, people looking to buy the ideal router for LEDE and the answer keeps shifting. You really need to be lucky to track down the right router that's available in the market and that there's a stable, good performing LEDE version for it. That happens at the low end, at the high end if you go x86, these concerns are alleviated. So, from an OpenWRT, effort focus and savings from distractions, I think the case can be made.
Now, on the market side. If you're a service provider, you want software that gives you hardware vendor lock-in avoidance and flexibility to develop new innovative services in the future. You don't want low quality software/firmware running on your edge hardware and spending your money on hardware replacement. CSPs are tired of that. Look at SDN, people today buy standard servers from HP, Dell, etc. and load them with SDN software that performs network functions, from firewalls, to edge routers, to more core functions. The equivalent is true on the home router side. When you look at the role of OpenWRT in that context and the associated economics, there's a clear business case both on the cost and revenue side.
Why? What is the compelling business reason behind that? As an ISP, the vast majority of my customers buy whatever I want them to, and I bill them monthly for it, even long after it has been fully paid off. It's a great revenue stream for me. Have problems? Great, buy a new modem from me. Why should I deal with software than end users can change? I've already got a back door into their CPE and already push config files and vendor-supplied updates to it.
Which will be the show stopper when it comes to financing (bank, investors, etc.)
Guys, @jeff, @jwoods, I believe we can agree to disagree. Depends on each one's experiences. This post is not about building a business plan or justifying the need to the forum users. This is a thought exercise that might or might not have value but it's intriguing to me and might be to others ... and you should respect that! To the extent you think the discussion is "pointless", please don't waste your valuable time on this.
Please don't waste others time asking the same questions over and over, and expecting a different answer.
I will be shutting down and deleting this discussion after you guys have the chance to reply. I'm sorry to see this discussion going down this path. Very disappointed.
To me that's the biggest thing right there. Who is the target market, and what do they want the hardware to do.
It seems to me there are two main markets of interest:
- everyday joe just wants a router with regular security updates, basically is OK with LEDE out of the box plus set up his own SSID
- People who actually have technical requirements
group 1 buys TP-LINK or the like, and it already works ok
group 2 would be best off with an espressobin or something like it or an x86 and separate access points.
group 2 is where large commercial entities aren't yet already satisfying the real needs
but group 1 is the group you seem to focus on... and it seems pointless compared to just flashing LEDE on more C7s
That does not seem to be aligned with the spirit of this forum, why not just stop posting to this thread but keep it for future reference?
1 Like