* [Cake] quad core arm
@ 2017-12-03 17:44 Dave Taht
2017-12-03 18:18 ` [Cake] [Cerowrt-devel] " Joel Wirāmu Pauling
` (2 more replies)
0 siblings, 3 replies; 6+ messages in thread
From: Dave Taht @ 2017-12-03 17:44 UTC (permalink / raw)
To: Cake List, cerowrt-devel
I have gone through a lot of hackerboards in the last few years.
I had multiple goals for them - primarily I wanted cheap and
cheap-to-power wifi and ethernet test targets. I settled on the
c.h.i.p. for a while for wifi.
Another goal was a largely fruitless quest to find the ideal next gen
replacement for the wndr3800. These days I'm using a AC2600 as my main
device and waiting for the ath10k support to catch up. I used to use
an olimex something or other for my NAS, I upgraded it to a pine64,
which was better but crashed hard a few months ago and I've not had a
chance to go fix it.
In most cases getting a modern kernel was a major problem. The odroid
C2 was my fastest network test target (can't drive 1gbit bidir tho),
and stuck on linux 3.10 for several years now. (I've heard rumors 4.x
almost works now).
Anyway the nanopi folk are now producing a wide range of boards I
haven't tried... til tomorrow:
https://www.amazon.com/gp/product/B0728LPB2R/ref=oh_aui_detailpage_o00_s01?ie=UTF8&psc=1
These appear to be supported on modern kernels in armbian (which has
thus far been the "best" distro for these hackerboards for me).
https://www.armbian.com/nanopi-neo-2/
On the really high end the 48 core arm boxes from cavium look interesting.
Anything else worth looking at?
--
Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Cake] [Cerowrt-devel] quad core arm
2017-12-03 17:44 [Cake] quad core arm Dave Taht
@ 2017-12-03 18:18 ` Joel Wirāmu Pauling
2017-12-03 19:49 ` [Cake] " Dave Taht
2017-12-04 0:11 ` Michael Richardson
2 siblings, 0 replies; 6+ messages in thread
From: Joel Wirāmu Pauling @ 2017-12-03 18:18 UTC (permalink / raw)
To: Dave Taht; +Cc: Cake List, cerowrt-devel
I quite liked the rk3399 board ; it has enough lanes for good duplex
ethernet support, but I haven't seen anyone bundling with multiple
Gigabit ports on a board so you are stuck with miniPciE add-in or USB3
dongles for more than one port.
You can get hold of them relatively easily via aliexpress and there
was a crowdfunded 'tinker' board variant out
:http://wiki.t-firefly.com/index.php/Firefly-RK3399/en
In addition to the rockchip board any of the sunxi - allwinner boards
have had a lot of mainline support added over the last 6 months. Most
of the easily available boards are all geared towards HTPC
applications however.
My current interest is in the c3xxx platform which FINALLY has enough
oomph to do duplex 20gbit and can be had with SFP+ equipped variants.
Lately most of the testing work I have been doing has been with the
thunderbolt3 network driver in net-next (and which I think is tagged
for inclusion with 4.15) which gives you 10-40gbit networking mostly
hasslefree.
-Joel
On 4 December 2017 at 06:44, Dave Taht <dave.taht@gmail.com> wrote:
> I have gone through a lot of hackerboards in the last few years.
>
> I had multiple goals for them - primarily I wanted cheap and
> cheap-to-power wifi and ethernet test targets. I settled on the
> c.h.i.p. for a while for wifi.
>
> Another goal was a largely fruitless quest to find the ideal next gen
> replacement for the wndr3800. These days I'm using a AC2600 as my main
> device and waiting for the ath10k support to catch up. I used to use
> an olimex something or other for my NAS, I upgraded it to a pine64,
> which was better but crashed hard a few months ago and I've not had a
> chance to go fix it.
>
> In most cases getting a modern kernel was a major problem. The odroid
> C2 was my fastest network test target (can't drive 1gbit bidir tho),
> and stuck on linux 3.10 for several years now. (I've heard rumors 4.x
> almost works now).
>
> Anyway the nanopi folk are now producing a wide range of boards I
> haven't tried... til tomorrow:
>
> https://www.amazon.com/gp/product/B0728LPB2R/ref=oh_aui_detailpage_o00_s01?ie=UTF8&psc=1
>
> These appear to be supported on modern kernels in armbian (which has
> thus far been the "best" distro for these hackerboards for me).
>
> https://www.armbian.com/nanopi-neo-2/
>
> On the really high end the 48 core arm boxes from cavium look interesting.
>
> Anything else worth looking at?
> --
>
> Dave Täht
> CEO, TekLibre, LLC
> http://www.teklibre.com
> Tel: 1-669-226-2619
> _______________________________________________
> Cerowrt-devel mailing list
> Cerowrt-devel@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cerowrt-devel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Cake] quad core arm
2017-12-03 17:44 [Cake] quad core arm Dave Taht
2017-12-03 18:18 ` [Cake] [Cerowrt-devel] " Joel Wirāmu Pauling
@ 2017-12-03 19:49 ` Dave Taht
2017-12-04 0:19 ` [Cake] [Cerowrt-devel] " Michael Richardson
2017-12-04 0:11 ` Michael Richardson
2 siblings, 1 reply; 6+ messages in thread
From: Dave Taht @ 2017-12-03 19:49 UTC (permalink / raw)
To: Cake List, cerowrt-devel
https://www.solid-run.com/marvell-armada-family/armada-8040-community-board/
looks rather promising. (recommendation courtesy koen koi)
I also picked up two 30 dollar 10GigE interfaces for spaceheater and
whatever I end up calling the second box
https://www.amazon.com/gp/product/B016OYD0D4/ref=oh_aui_detailpage_o02_s00?ie=UTF8&psc=1
I guess in part what I'm mentally shifting to, is what happens after
cake? I'd rather like to have something an ISP could use on or near a
head end, and that entails moving stuff into hardware (and a few more
O(1) algorithms).
Mellonox at least used to make a board with hackable FPGAs, and
perhaps marvell will be more open about how to program the offload
engine next time, and P4 is coming along.
On Sun, Dec 3, 2017 at 9:44 AM, Dave Taht <dave.taht@gmail.com> wrote:
> I have gone through a lot of hackerboards in the last few years.
>
> I had multiple goals for them - primarily I wanted cheap and
> cheap-to-power wifi and ethernet test targets. I settled on the
> c.h.i.p. for a while for wifi.
>
> Another goal was a largely fruitless quest to find the ideal next gen
> replacement for the wndr3800. These days I'm using a AC2600 as my main
> device and waiting for the ath10k support to catch up. I used to use
> an olimex something or other for my NAS, I upgraded it to a pine64,
> which was better but crashed hard a few months ago and I've not had a
> chance to go fix it.
>
> In most cases getting a modern kernel was a major problem. The odroid
> C2 was my fastest network test target (can't drive 1gbit bidir tho),
> and stuck on linux 3.10 for several years now. (I've heard rumors 4.x
> almost works now).
>
> Anyway the nanopi folk are now producing a wide range of boards I
> haven't tried... til tomorrow:
>
> https://www.amazon.com/gp/product/B0728LPB2R/ref=oh_aui_detailpage_o00_s01?ie=UTF8&psc=1
>
> These appear to be supported on modern kernels in armbian (which has
> thus far been the "best" distro for these hackerboards for me).
>
> https://www.armbian.com/nanopi-neo-2/
>
> On the really high end the 48 core arm boxes from cavium look interesting.
>
> Anything else worth looking at?
> --
>
> Dave Täht
> CEO, TekLibre, LLC
> http://www.teklibre.com
> Tel: 1-669-226-2619
--
Dave Täht
CEO, TekLibre, LLC
http://www.teklibre.com
Tel: 1-669-226-2619
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Cake] [Cerowrt-devel] quad core arm
2017-12-03 17:44 [Cake] quad core arm Dave Taht
2017-12-03 18:18 ` [Cake] [Cerowrt-devel] " Joel Wirāmu Pauling
2017-12-03 19:49 ` [Cake] " Dave Taht
@ 2017-12-04 0:11 ` Michael Richardson
2017-12-04 0:34 ` Joel Wirāmu Pauling
2 siblings, 1 reply; 6+ messages in thread
From: Michael Richardson @ 2017-12-04 0:11 UTC (permalink / raw)
To: Dave Taht; +Cc: Cake List, cerowrt-devel
[-- Attachment #1: Type: text/plain, Size: 2275 bytes --]
Dave Taht <dave.taht@gmail.com> wrote:
> I had multiple goals for them - primarily I wanted cheap and
> cheap-to-power wifi and ethernet test targets. I settled on the
> c.h.i.p. for a while for wifi.
I have a few (3-4) that I'm not using if you want them.
(Never removed from the box)
> In most cases getting a modern kernel was a major problem. The odroid
> C2 was my fastest network test target (can't drive 1gbit bidir tho),
> and stuck on linux 3.10 for several years now. (I've heard rumors 4.x
> almost works now).
I didn't think that odroid had problems taking the latest kernels.
I just bought one for a project involving hubofallthings.
> Anyway the nanopi folk are now producing a wide range of boards I
> haven't tried... til tomorrow:
> https://www.amazon.com/gp/product/B0728LPB2R/ref=oh_aui_detailpage_o00_s01?ie=UTF8&psc=1
> These appear to be supported on modern kernels in armbian (which has
> thus far been the "best" distro for these hackerboards for me).
> https://www.armbian.com/nanopi-neo-2/
I'm using the Orange PI Zero in a project, and armbian debian has 4.13.16
kernel. But, I'm not trying to push gigabit ethernet; rather the opposite.
I just want wired connectivity. Sadly the "PoE" support for the OPI0 is not
IEEE spec, just a jumper that lets you feed 5V in... Rumours are that the
built-in wifi sucks, but I actively don't want wifi anyway. I do like that
it can boot uboot from NOR FLASH.
I believe I will 3D print a case that accomodates two USB interfaced
ethernets via pins on the system, so that from the outside it looks smart.
I will have to pick my USB devices carefully as I need auto-X PHYs, which
usually only comes with GbE.
> On the really high end the 48 core arm boxes from cavium look interesting.
I'm told that there is some special sauce to get them to go at the speeds the
specs say. Basically you run Linux on one core, and a special (binary blob)
micro kernel on the other 47, and manage them via FORCES.
--
] Never tell me the odds! | ipv6 mesh networks [
] Michael Richardson, Sandelman Software Works | network architect [
] mcr@sandelman.ca http://www.sandelman.ca/ | ruby on rails [
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Cake] [Cerowrt-devel] quad core arm
2017-12-03 19:49 ` [Cake] " Dave Taht
@ 2017-12-04 0:19 ` Michael Richardson
0 siblings, 0 replies; 6+ messages in thread
From: Michael Richardson @ 2017-12-04 0:19 UTC (permalink / raw)
To: Dave Taht; +Cc: Cake List, cerowrt-devel
[-- Attachment #1: Type: text/plain, Size: 1547 bytes --]
Dave Taht <dave.taht@gmail.com> wrote:
> https://www.solid-run.com/marvell-armada-family/armada-8040-community-board/
> looks rather promising. (recommendation courtesy koen koi)
I paid for one of these back in early April to do a variety of things with that
could benefit from offload. It was supposed to ship at the end of April.
After ignoring me repeatedly, they promised in August it would ship by the
end of September once they received their Marvell parts.
I pushed them to tell me what date they had been promised Marvell parts, and
they wouldn't say. At the end of September no boards had been fabricated and
I asked for and received a refund.
I would buy it again as it looks like a nice evaluation platform for a bunch
of things, but for the moment I have no time.
Maybe it's shipping now.
> I guess in part what I'm mentally shifting to, is what happens after
> cake? I'd rather like to have something an ISP could use on or near a
> head end, and that entails moving stuff into hardware (and a few more
> O(1) algorithms).
> Mellonox at least used to make a board with hackable FPGAs, and
> perhaps marvell will be more open about how to program the offload
> engine next time, and P4 is coming along.
So my use was for an ISP edge (PPPoE BMS).
--
] Never tell me the odds! | ipv6 mesh networks [
] Michael Richardson, Sandelman Software Works | network architect [
] mcr@sandelman.ca http://www.sandelman.ca/ | ruby on rails [
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [Cake] [Cerowrt-devel] quad core arm
2017-12-04 0:11 ` Michael Richardson
@ 2017-12-04 0:34 ` Joel Wirāmu Pauling
0 siblings, 0 replies; 6+ messages in thread
From: Joel Wirāmu Pauling @ 2017-12-04 0:34 UTC (permalink / raw)
To: Michael Richardson; +Cc: Dave Taht, Cake List, cerowrt-devel
On 4 December 2017 at 13:11, Michael Richardson <mcr@sandelman.ca> wrote:
>
> Dave Taht <dave.taht@gmail.com> wrote:
.
>
> > On the really high end the 48 core arm boxes from cavium look interesting.
>
> I'm told that there is some special sauce to get them to go at the speeds the
> specs say. Basically you run Linux on one core, and a special (binary blob)
> micro kernel on the other 47, and manage them via FORCES.
>
Can confrim; we included a Cavium GPC in our Original VSG7850 as a
Linux Environment along side the Trident2+ chipset.
It was a nightmare just getting the NDK signed to get access to the
Cavium build chain, and then it was using an ancient 2.6 kernel with
Secret sauce patches.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2017-12-04 0:35 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-12-03 17:44 [Cake] quad core arm Dave Taht
2017-12-03 18:18 ` [Cake] [Cerowrt-devel] " Joel Wirāmu Pauling
2017-12-03 19:49 ` [Cake] " Dave Taht
2017-12-04 0:19 ` [Cake] [Cerowrt-devel] " Michael Richardson
2017-12-04 0:11 ` Michael Richardson
2017-12-04 0:34 ` Joel Wirāmu Pauling
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox