Announcing OpenWrt/MLPPP - multilink firmware for consumer routers - Caneris & Acanac

[Lennart Sorensen]

On Tue, Mar 02, 2010 at 01:53:03PM -0500, D. Hugh Redelmeier wrote:
> | From: Lennart Sorensen <lsorense-1wCw9BSqJbv44Nm34jS7GywD8/FfD2ys at public.gmane.org>
> 
> | Actually some DIR-825 rev B seem to have 64MB according to the dd-wrt
> | developers.  Some have 32MB.  No one seems to know why this is.
> 
> Nice!  It would be nice to know if you could some how tell before you
> buy.

None that I know of.  It seems dlink considers them equivelant and it's
just a build variation.  Not sure if they need a different FCC id if
they change the ram like that, so perhaps there is a difference.

> I forgot to mention another interesting cheap choice: the TP-Link
> TL-WR1043ND.  It is available several places, but this looks to be the
> cheapest local source at the moment ($52.99 with free delivery in
> the GTA):
>   <http://www.bewawa.com/tp-link-ultimate-wireless-n-300mbps-gigabit-router-w-3-detachable-antennas-wr1043nd.html>
> 
> This seems to be:
> - all atheros
> - slower CPU (only 400MHz, not 600MHz like the others we were
>   discussing)
> - 1G WAN and LAN ethernet
> - one USB 2.0 port
> - 32M RAM, 8M flash
> - support in OpenWRT tree
>   <http://wiki.openwrt.org/inbox/tp-link.tl-wr1043nd>

Interesting little device.

> I have no experience with this router or even with this brand.  (I was
> going to say "manufacturer", but brands and manufacturers are not
> isomorphic.)  I read a comment somewhere that TP-Link makes low cost
> knock-offs of other companies products so there is a lack of coherence
> to their product line as a whole.

Knock-offs including copying the software?

> How much does CPU crunch matter in these devices?
> 
> Some people have said that routers with USB ports make bad file
> servers because USB takes considerable CPU and router CPUs are not up
> to saturating a drive -- I don't know if this is true.  Some routers
> (Asus) have defective USB ports that don't even support high speed USB
> (contrary to their spec sheets).

USB is very cpu intensive, so yes that maeks sense.

> I don't imagine that the CPU is involved directly in the switching of
> packets.  If it were, then 1G ethernet might put new loads on the CPU.
> Actually routing the packet surely does involve the CPU so the faster
> CPU may be useful there.  Maybe the new ADSL2+ service could make the
> faster CPU worthwhile.

With a decent network interface, a 400MHz or so CPU should be able to
do a good chunk of a gigabit link.  It might even manage it.  It depends
on many factors though.

> Bulk crypto should be offloaded from the CPU.  It is in at least some
> of these consumer routers but I'm not sure which and I'm not sure
> which crypto hardware is available to open source.  If the CPU is
> doing crypto, the faster CPU would be helpful.

Many crypto engines have linux kernel support.  not all, but many do.
If present and supported they tend to help a lot.

> Without measurement, guesses about CPU requrements are just that.

Yep.

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists