dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
20948

Anav
Sarcastic Llama? Naw, Just Acerbic
Premium Member
join:2001-07-16
Dartmouth, NS

Anav to lorennerol

Premium Member

to lorennerol

Re: USG 100 WAN SPEED

said by lorennerol:

It's perhaps worth noting that we've out 50-75 ZyXEL routers into the field over the last 15 years. Probably 25 are still in service. Of those no longer in service, ONE failed. Every other one was retired because it was too slow for an upgraded Internet connection.

But telling a client they need to replace a $200-$300 ZyWALL 10 or 10+ that's been in service for six years is a different conversation that having to replace a $600 USG 100 that's been in service less than a year because it can't keep up with their $100/month 50/10 Internet connection. What's happening is that the ZyXEL's model of pricing routers based on throughput isn't working now that Comcast and other ISPs are offering blistering fast (100/20 for $190/month) connections for business customers. Maybe they need to look at a different tiering/pricing model.

Well it wouldn't be so bad if it met the specs (100-20 should be doable by a 225 speced router. zyxel USA should recognize this and offer a decent upgrade path. A usg 100 is $370ish at provantage not $600 (oops sorry you told your clients they were $600 or was that $230 to configure LOL
JPedroT
Premium Member
join:2005-02-18

JPedroT to mozerd

Premium Member

to mozerd
Price/Performance might not be what the people expect/want, but are there any other vendors out there that are comparative in the feature set and is on a much lower price?
lorennerol
Premium Member
join:2003-10-29
Seattle, WA

lorennerol

Premium Member

said by JPedroT:

Price/Performance might not be what the people expect/want, but are there any other vendors out there that are comparative in the feature set and is on a much lower price?

That's the thing; we don't use 90% of the features. Maybe they need to move to a unified hardware platform (which the bean counters would like) and make the tiers based on software features. Basically, restrict the performance/feature set in software. Buy a USG110 now, but if you need more speed, upgrade it to a 210 later.

As-is, I'm starting to look at (free) open source options that can actually handle QoS and BWM for VoIP correctly.
FirebirdTN
join:2012-12-13
Brighton, TN

FirebirdTN to mozerd

Member

to mozerd
I'm a little late getting in on the conversation here, but I just don't believe a USG 100 can't handle 100/20 service.

There has got to either be something wrong with the OP's router, or some weird incompatibility.

I have 50/10 at the house (with powerboost), and my USG50 has NO problem with it. Now I have enabled BWM to defeat powerboost, BUT...before I defeated it I would peak out at around 105-110 with PC direct to the modem, and about 95-97 through the router.

I haven't done any real testing, but I suspect the USG 50 is only good until about 100. So the USG 100 has to be good for at least that.

I do have firewall and ADP active, but no UTM services (No AV, IDP, or content filtering).

-Alan

Brano
I hate Vogons
MVM
join:2002-06-25
Burlington, ON
(Software) OPNsense
Ubiquiti UniFi UAP-AC-PRO
Ubiquiti NanoBeam M5 16

Brano to mozerd

MVM

to mozerd
I'm also surprised that the speeds mozerd See Profile is getting are asymetrical. When I did my tests here »USG200 speed tests #3 I've tested it both ways with same results.

Anav
Sarcastic Llama? Naw, Just Acerbic
Premium Member
join:2001-07-16
Dartmouth, NS

1 edit

Anav to mozerd

Premium Member

to mozerd
Did some JPerf 2.0.2 tests.
IMAC 2008 and windows laptop circa 2011.

Through (connected to) same GIG switch
1a Results IMAC client...... 620Mbps
1b Results IMAC Server 220Mbps

Through home network with laptop connected to lan port on usg300
2a Results IMAC client .....620Mbps
2b Results IMAC Server.....220Mbps

Not sure why the results differ so greatly from laptop to IMAC depending if one is a server or not??? In any case did show that gig network is at least somewhat working and a valid thruput environment to test USG100 through its WAN port.

USG100 connected via wan port to lan port of usg300
Laptop connected to LAN port of USG100
3a. Results IMAC client.......... 82Mbps
3b. Results IMAC Server........ 92Mbps

Thus USG100 on WAN side is not performing close to expectations. I would have expected one way TX in the order of 180-200Mbps.

Redid the tests above to confirm throughput.
Redid without firewall on, on the USG100.

4a. Results IMAC Client...........103Mbps
4b. Results IMAC server.......... 115Mbps
lorennerol
Premium Member
join:2003-10-29
Seattle, WA

lorennerol to mozerd

Premium Member

to mozerd
We currently have three clients on 100/20 Comcast circuits. Two have USG200 routers and are testing out (using Speedtest.net) at 110/22 or so.

The one on a USG100 is consistently testing around 51/8.

And I called Comcast to confirm that they have the correct config on the modem. I'll test directly through the modem when I'm on-site later this week.

ESA13
@rogers.com

ESA13

Anon

I can confirm this problem with the USG100. I have two sites that have consistently only got 57% of there through put (23Mbs of the 40Mbs DL). Once the USG100 was pulled from the link all speeds hit there respected 40Mbs of higher. I have changes IPS modems gone through so many reconfigurations with them. It makes no sense since the USG100 never has a high CPU or memory usage. Very frustrating. This looks to be a sad case of manufacturer issues. I have the newest firmware on both. But this issue was even on V3.

Anav
Sarcastic Llama? Naw, Just Acerbic
Premium Member
join:2001-07-16
Dartmouth, NS

Anav to mozerd

Premium Member

to mozerd
To keep this thread going, I am next going to use JPERF again (Iperf for dummies ie graphical window version) except I want to use UDP traffic instead.

The problem is I must be doing something wrong. I cannot get any results that make sense.
Using TCP I can get reasonable results on my network (old IMAC on home network through long cables and two gigabit switches to a pc laptop fairly new attached directly to lan port on USG300. 220Mbps from PC(client) to IMAC(server) or 630Mbps from IMac (client) to PC (server). So my environment is reasonable to test the theoertical one way 225Mbps of the USG100.

Now I expect those test are under ideal conditions so I reasonably expect real world results in the 180Mbps area.

problem is no matter what setup up I use for UDP I get whacky incosistent results such as 1Mbps up to I think 30Mbps. So either I am not doing something correctly or UDP doesnt work on my home network. I think the former is more likely (being an amateur tester).

Thus need help with JPerf 2.02. to get useful UDP throughput results for my testing. There is no point hooking up the USG100 to the 300 (wan port to lan port) until i can get home network results that make sense.

I remembered that I used ixia's Qcheck in the past but although I can get qcheck for the windows PC, I cannot find it for the MAC OS X. I did download a MAC OS X endpoint application but it asked for password. :-(
Kirby Smith
join:2001-01-26
Derry, NH

Kirby Smith to mozerd

Member

to mozerd
What we are missing in this thread to give it the lifetime you want, as well as provide useful SOHO information, is an ISA570 comparison with the USG100 in terms of Mbps of real throughput (using some ISP connection) per unit cost (at Provantage or wherever). If the ISA570 (with its higher absolute speeds) can do better in price performance ratio, then small businesses (and overkill torrent users ) wanting to keep up with ISP speed increases should move in that direction. This assumes that Mozerd hasn't found a flaw since the rave on the big USG improvements thread.

k

Anav
Sarcastic Llama? Naw, Just Acerbic
Premium Member
join:2001-07-16
Dartmouth, NS

1 edit

Anav

Premium Member

I have no clue how ISA570 will have an iota of impact on my life nor in this forum suggest perhaps you should be posting here......
»Other Manufacturers. oops that not an edgemax product but a crisco one, different forum LOL

Looking for Iperf acumen at the moment.

Brano
I hate Vogons
MVM
join:2002-06-25
Burlington, ON
(Software) OPNsense
Ubiquiti UniFi UAP-AC-PRO
Ubiquiti NanoBeam M5 16

Brano to Anav

MVM

to Anav
said by Anav:

To keep this thread going, I am next going to use JPERF again (Iperf for dummies ie graphical window version) except I want to use UDP traffic instead.

On the product brief near to the performance is stated
quote:
Maximum throughput based on RFC 2544 (1,518-byte UDP packets).

lorennerol
Premium Member
join:2003-10-29
Seattle, WA

lorennerol to Anav

Premium Member

to Anav
said by Anav:

I have no clue how ISA570 will have an iota of impact on my life

Since you resell ZyXEL gear, it will impact your life if it has the throughput performance that the USG routers don't (but should have).

At this point, new routers from ZyXEL with anything less than Gigabit throughput on the WAN interface with services running is not going to cut it. Google is already offers this speed and I suspect Comcast and others will be there in the next 3-5 years as well (well within the expected lifespan of a business class router).
JPedroT
Premium Member
join:2005-02-18

JPedroT

Premium Member

There is a big difference between 1Gb WAN port and 1Gb throughput.
Most of the time, your PC wont be able to push or receive that amount of data anyway. IE read/write speeds of HDD/SSD impacts quite a lot.
But if we are talking multiple hosts then its a different matter.

Anav
Sarcastic Llama? Naw, Just Acerbic
Premium Member
join:2001-07-16
Dartmouth, NS

Anav to lorennerol

Premium Member

to lorennerol
said by lorennerol:

said by Anav:

I have no clue how ISA570 will have an iota of impact on my life

Since you resell ZyXEL gear, it will impact your life if it has the throughput performance that the USG routers don't (but should have).

At this point, new routers from ZyXEL with anything less than Gigabit throughput on the WAN interface with services running is not going to cut it. Google is already offers this speed and I suspect Comcast and others will be there in the next 3-5 years as well (well within the expected lifespan of a business class router).

Did they start selling weed early in Wash. Cisco doesnt affect me at all. I am not selling much of anything these days dont have the energy.
Anav

Anav to Brano

Premium Member

to Brano
said by Brano:

said by Anav:

To keep this thread going, I am next going to use JPERF again (Iperf for dummies ie graphical window version) except I want to use UDP traffic instead.

On the product brief near to the performance is stated
quote:
Maximum throughput based on RFC 2544 (1,518-byte UDP packets).

Ahh okay, that helps a little. Hmm must use UDP to inflate their numbers as UDP doesnt verify reception. Do all vendors provide garbage numbers using UDP packets?
lorennerol
Premium Member
join:2003-10-29
Seattle, WA

lorennerol to JPedroT

Premium Member

to JPedroT
said by JPedroT:

But if we are talking multiple hosts then its a different matter.

These are expensive business-class routers; I don't think there are many out there sitting in front of single PCs. We have one client looking at 2x100/20 circuits from Comcast right now. And I don't think their one year old USG 200 will cut it.
lorennerol

lorennerol to Anav

Premium Member

to Anav
Yeah, and GM doesn't sell Toyotas, but Toyota sales sure affect GM.

PS- The idiocy of our existing pot law is that it's currently legal to possess and use small amounts, but illegal to produce, buy, or sell. So it's all good, as long as it magically materialized in your hand.
JPedroT
Premium Member
join:2005-02-18

JPedroT to lorennerol

Premium Member

to lorennerol
I have no problem with agreeing with you in this.
But then that is also how the performance test should be used then..

Because single session might not be able to get the full throughput out a device due to a PC not being capable enough or the device being multicore. And dividing a session across multiple cores does not implicitly mean the performance scales linearly with the number of cores you throw at it due to overhead for context switching and so forth.

So you might need multiple PC's to generate enough traffic and to get the max out of the horsepower in the device.

As for ZyXEL providing a device, sure they should and they probably will. Since they already got 1Gbps performance on DSL/Ethernet Gateways.
lorennerol
Premium Member
join:2003-10-29
Seattle, WA

lorennerol

Premium Member

I'm sitting, right now, in front of a 24 port ZyXEL PoE switch that will only run 12 phones, even though those 12 phones are only using 1/3 of the rated wattage. ZyXEL is utterly baffled and has sent me beta firmware and another switch, all with the exact same result.

Not feeling up about these guys right now.
JPedroT
Premium Member
join:2005-02-18

JPedroT

Premium Member

Thats because its a chipset limitation I guess, I have seen this on other switches also. Just that it was printed clearly on the datasheet for those other switches.
Which ZyXEL switch is it you have?
lorennerol
Premium Member
join:2003-10-29
Seattle, WA

lorennerol

Premium Member

said by JPedroT:

Thats because its a chipset limitation I guess, I have seen this on other switches also. Just that it was printed clearly on the datasheet for those other switches.
Which ZyXEL switch is it you have?

ES-2024. If it's a chipset limitation, someone should clue-in ZyXEL: They are convinced that it should support more than 12 devices and are selling it that way. FWIW, if I change it from Consumption mode to Allocation mode it only runs 11.
JPedroT
Premium Member
join:2005-02-18

JPedroT

Premium Member

said by lorennerol:

said by JPedroT:

Thats because its a chipset limitation I guess, I have seen this on other switches also. Just that it was printed clearly on the datasheet for those other switches.
Which ZyXEL switch is it you have?

ES-2024. If it's a chipset limitation, someone should clue-in ZyXEL: They are convinced that it should support more than 12 devices and are selling it that way. FWIW, if I change it from Consumption mode to Allocation mode it only runs 11.

I can not say for sure, but its my guess, based on similar experience with other switches

mozerd
Light Will Pierce The Darkness
MVM
join:2004-04-23
Nepean, ON

mozerd

MVM

I'm VERY dissapointed in the WAN/LAN/WAN metrics. A gig wan port is pure BS insofar ZyXEL and the USG100 is concerned. I've been snookered and it does not feel good.
JPedroT
Premium Member
join:2005-02-18

JPedroT

Premium Member

said by mozerd:

I'm VERY dissapointed in the WAN/LAN/WAN metrics. A gig wan port is pure BS insofar ZyXEL and the USG100 is concerned. I've been snookered and it does not feel good.

Where does it say that the gig wan port means gigabit throughput?
lorennerol
Premium Member
join:2003-10-29
Seattle, WA

lorennerol

Premium Member

said by JPedroT:

Where does it say that the gig wan port means gigabit throughput?

+1. I was certainly under no impression that I would get a Gigabit of Internet throughput from any ZyXEL router.

And amidst all the bashing I should say that I've only ever had ONE ZyXEL router fail (and that was a USG 200 last month), and I've probably put 50+ in the field over 10+ years, going back to the ZyWALL 10. Many have been retired due to throughput issues, but usually after many years of use. And I've routinely had them run 1-2 years without needing a reboot. They have been solid routers for us, they just need an order of magnitude better Internet throughput right now.

Anav
Sarcastic Llama? Naw, Just Acerbic
Premium Member
join:2001-07-16
Dartmouth, NS

Anav

Premium Member

+1
Anav

Anav to mozerd

Premium Member

to mozerd
Latest Update On throughput tests.
Firmware - LATEST just released Vers 3.30 (on both USG300 and 100)

Environment. IMAC attached to home network and through gigabit switches and cabling to LAN port of USG 300 the main router. PC laptop attached to LAN port of USG 100.
USG 100 Wan port connected to LAN port of USG 300. Throughput of the home network is above 200Mbps in both directions and is not a factor.

LAN to WAN Throughput: - PC (client) to IMAC (server)
TCP FW OFF - 195 Mbps
TCP FW ON - 145 Mbps

WAN to LAN Throughput: - IMAC (client) to PC (server)
TCP FW OFF - 185 Mbps
TCP FW ON - 140 Mbps

Assumption: Additional overall difference (loss) on Wan to LAN due to NAT.
Results: With a combination of a bit more experience using IPERF (JPERF Front end) and mostly due to using latest firmwares results are much better than previously reported. I still believe that the results that I received are lower than they should be with firewall on.
JPedroT
Premium Member
join:2005-02-18

JPedroT

Premium Member

Since you are using TCP, did you calculate your bandwidth delay product? What TCP window size are you using? Did you try to use parallel streams?

And you should also look at nuttcp, if you like to play with tests.

»www.nuttcp.net/nuttcp/We ··· age.html

Would be interesting to see a ftp test also, just install a ramdisk on the server and the client and do file transfer, preferably in the 500MB to 1GB range like an ISO file

Anav
Sarcastic Llama? Naw, Just Acerbic
Premium Member
join:2001-07-16
Dartmouth, NS

Anav

Premium Member

Hey JP, thanks for the help.
What is bandwidth delay product. I notice there is a checkbox for NO DELAY. What would this do and what is the effect of bandwidth delay? For TCP window size I ranged around 56bytes to 1500 bytes. What I noticed is that over about 40Bytes there was no appreciable difference in performance.

What should I expect to see??

There are no useful guides on testing and playing with the parameters and what effects different combinations will have. I suspect that most companies expect testers to be know what the heck they are doing not like stab in the dark amateurs like me.

Which brings me to UDP testing. That was a dogs breakfast, by playing with parameters UDP bandwidth size I seemed to be able to have no limits on throughput, ie up 400Mbps even which I know is ludicrous so I must be doing something fundamentally wrong, but cannot find any guidance or explanation or assistance.

Thus I stuck with TCP and basic settings. Parallel streams to test what (a synthetic dual up and down scenario??) I will try a few more combos if you like.