WAN to LAN: how much is enough?

[fusspot]

Hello - I am seeking case-specific guidance on choosing a main router to sit between my cable modem and home network, with the most important consideration between WAN to LAN throughput. That said, I don't want to simply pick the highest-rated (and likely pricey) unit from your chart if it is overkill for my real world needs. In other words, how much WAN to LAN is enough for this scenario...

Here is a description of my current pipeline:

1) Comcast Seattle Xfinity Internet Performance package offering up to 20 Mbps downlink. Modem is a Surfboard 6120. Connected by CAT-5 to...

2) Linksys WRT54GS acting as main router to distribute DHCP to network. Wireless mode is turned off. Connected by CAT-5 to...

3) Eight port ethernet hub which distributes CAT-5 to eight rooms in the house, of which only three are actively used, with typically not more than two in use at once.

4) Room One connects via CAT-5 to an Apple AirPort Express (early 'g' model) for the purpose of using AirPlay with the stereo system. Wireless mode is turned off.

5) Room Two connects via CAT-5 to an Apple Mac Pro.

6) Port Three connects via CAT-5 to an Apple AirPort Extreme (newest, 5th Gen) operating in Bridge Mode. Downstream from the AirPort are two CAT-5 connected devices: A) an Apple MacBook Pro, and B) a Vonage V-Portal adapter. The AirPort also acts as a print server via the USB port. Lastly, the AirPort provides a wireless signal to the home.

Here's the problem... I have been using Speedtest.net to do some performance testing. I recognize from articles on this forum that this may not be the best way, but I think it is sufficiently apples-to-apples in order to make valid the following observations:

When I test from one or the other computer, I consistently range from mid-3s Mbps to low-4s Mbps. (Note - when testing, all other devices are physically disconnected/disabled so that I am not adding any bandwidth degradation.)

Now, if I connect the modem directly to the ethernet hub and perform the same tests, the speed consistently doubles, ranging from low-7s Mbps to low-8s Mbps.

I have connected the MacBook Pro directly to the modem via Cat-5, and also tried placing the AirPortExtreme in between the those two. In both scenarios, I get comparable results -- performance doubles -- as to when the hub is in-line. So from that I conclude the degradation is caused by the WAN to LAN throughput of the Linksys acting as main router, and not an issue with my hub or cabling (no runs are anywhere close to 100m).

So, assuming that A) my cable service is choking on neighborhood bandwidth limits far below the promised "up to 20 Mbps," and 2) it ain't gonna get any better as more people come to the neighborhood, and 3) I'm not going to pay Comcast for a higher-tier package...

I need a new main router, and wonder where would be my sweet spot on the WAN to LAN chart? Should I go for the 600+ monsters? Would something in the 400s be adequate, noting that the AirPort Extreme is in that range and performed fine in my network? Or are there some totally different factors that I'm ignorant of and need to consider?

Lastly, please note that hardware locations and user functions require replacement of the Linksys in its current position in-line, as opposed to a complete reorganization of the other existing equipment.

Thanks in advance for your kind assistance. Your website has been very useful to me in contemplating this question, and I'll value any expert opinions that can help me resolve this need. Cheers!

 

[kanewolf]

Are you using Comcast's speedtest or one of the generic ones? http://speedtest.comcast.net/
You may see your "advertised" bandwidth testing to their internal speedtest portal. I know that with FIOS, it is iffy to get full bandwidth to the generic speedtest destinations but I always see my advertised bandwidth to the FIOS internal speedtest portal.

 

[fusspot]

Kanewolf - I'm using Speedtest.net, which is third-party. I'm able to select a variety of testing servers in the region, all of which provide essentially the same results. Thanks.

 

[sfx2000]

The bottleneck is the WRT54GS acting as the WAN entry point. It was a good router "in the day", but the performance of the actual SoC has been eclipsed by newer chipsets and software. Newer routers are more efficient, and have better NAT management for more clients.

One thing to perhaps consider, is picking up an E2500 or E3200 from Linksys refurb store, disable the wireless, and drop it in place of the WRT54G. Routing performance on those routers is decent... if cost is no object, pick up another AP Extreme and a GIGe switch to extend the ports and expand the wireless footprint.

 

[fusspot]

sfx2000 - Thanks for the reply; a couple of follow-ups, please... The two Linksys refurbs you noted, I don't see those in the WAN-to-LAN performance grid. Any idea where they stack up in that mix? Also, could you please explain you last sentence about using another AP and a gigE switch. What would be the benefits of that? Thank you.

 

[sfx2000]

sfx2000 - Thanks for the reply; a couple of follow-ups, please... The two Linksys refurbs you noted, I don't see those in the WAN-to-LAN performance grid. Any idea where they stack up in that mix? Also, could you please explain you last sentence about using another AP and a gigE switch. What would be the benefits of that? Thank you.

They're mid-pack, better than the WRT54G w/DD-WRT - like I mentioned earlier, the WRT54G is a good router, just showing its age - I still like them though, extremely hackable - and with OpenWRT, very useful as access points.

AP == Airport Extreme - it has good routing performance - the GIGe switch was to add additional ports, as the Airport is limited to 3 external LAN ports, probably the weakest aspect of that platform.

 

[thiggins]

Both the E2500 and E3200 are in the Router Charts.

All you need is a router with routing throughput a bit higher (5-10% for some headroom) than the bandwidth your ISP provisions.