PDA

View Full Version : 50M Power levels + channel bonding (not an issue just curious)


TJS
29-04-2011, 17:30
I've seen on here some people on 50 and even 30 who bond 4 channels, does this make any difference?

also how are these power levels?


Startup Procedure
Procedure Status Comment
Acquire Downstream Channel 298750000 Hz Locked
Connectivity State OK Operational
Boot State OK Operational
Configuration File OK
Security Enabled BPI+
Downstream Channels
Lock Status Modulation Channel ID Max Raw Bit Rate Frequency Power SNR Docsis/EuroDocsis locked
Locked QAM256 104 55616000 Kbits/sec 298750000 Hz 5.1 dBmV 33.6 dB Hybrid
Locked QAM256 105 55616000 Kbits/sec 306750000 Hz 5.3 dBmV 33.6 dB Hybrid
Locked QAM256 106 55616000 Kbits/sec 314750000 Hz 5.5 dBmV 33.5 dB Hybrid
Unlocked Unknown 0 0 Ksym/sec 0 Hz 0.0 dBmV 0.0 dB Unknown
Unlocked Unknown 0 0 Ksym/sec 0 Hz 0.0 dBmV 0.0 dB Unknown
Unlocked Unknown 0 0 Ksym/sec 0 Hz 0.0 dBmV 0.0 dB Unknown
Unlocked Unknown 0 0 Ksym/sec 0 Hz 0.0 dBmV 0.0 dB Unknown
Unlocked Unknown 0 0 Ksym/sec 0 Hz 0.0 dBmV 0.0 dB Unknown
Upstream Channels
Lock Status Modulation Channel ID Max Raw Bit Rate Frequency Power
Locked ATDMA 3 20480 Kbits/sec 45800000 Hz 41.7 dBmV
Unlocked Unknown 0 0 Ksym/sec 0 Hz 0.0 dBmV
Unlocked Unknown 0 0 Ksym/sec 0 Hz 0.0 dBmV
Unlocked Unknown 0 0 Ksym/sec 0 Hz 0.0 dBmV

the power levels on the downstream go up to near 7 dBmV at night (not sure why) but there are no errors at all in the event log and i have been getting stable 50 constantly

---------- Post added at 16:30 ---------- Previous post was at 16:30 ----------

Also what kind of UBR will i be on a motorola or a cisco, and which is better?

pip08456
29-04-2011, 17:33
Depends if you are on a UBR or BSR as to the number of channels you have so nothing to worry about.

TJS
29-04-2011, 17:35
what is a BSR, also the box inside my house that the cable comes out of says CABLE&Wireless on it if that means anything? but this house has been cable connected for a good many years, one of the phone sockets says NYNEX on it

General Maximus
29-04-2011, 18:01
it is just the company stamp for whoever owned the hardware at the time. You could have something installed now which says Virgin Media on it and they might get taken over by another company in 10 years time.

Power levels look good btw

pip08456
29-04-2011, 18:15
UBR - Universal Broadband Router (Cisco)

BSR - Bootstrap Router (Motorola)

Just different makes of CMTS (Cable Modem Termination Server), hence the difference in number of channels.

General Maximus
29-04-2011, 18:22
hence the difference in number of channels.

will the difference in the number of channels they are able to provide ultimately create a cap on the throughput isps are able to provide as there won't be as many channels to bond?

Ignitionnet
29-04-2011, 18:34
BSR - Bootstrap Router (Motorola)

Broadband Services Router.

pip08456
29-04-2011, 18:36
Oooops!

Ignitionnet
29-04-2011, 18:36
will the difference in the number of channels they are able to provide ultimately create a cap on the throughput isps are able to provide as there won't be as many channels to bond?

The difference between 3 and 4 channels is purely down to VM's implementation. Ultimately there will be a cap on throughput, which is when you upgrade the line cards.

Mainstream modems can bond 8 channels, the existing hardware on the BSRs, the new line cards and the M-CMTS hardware on the uBR are all happily capable.

General Maximus
29-04-2011, 22:02
so what are we looking at?

8 x 55mbit = 440mbit absolute max on current hardware?

jb66
29-04-2011, 22:22
In theory but it will be hard to max out 8 channels, they might bring out a 16 channel modem with 25 on each = 400meg

Ignitionnet
30-04-2011, 00:41
The Pooper Hub will do 400Mbit/s.
The VMNG300 will do 200Mbit/s.
The VM256 will do 50Mbit/s.

In short it's a non-issue. The hardware is well ahead of the tiers being offered and will continue to be so.

General Maximus
30-04-2011, 09:42
it is such a damn shame that the VMNG can do 200mbit and they wont even give it the chance to handle half of that. If it has got some much longevity in it you would have thought they would want to swap out all the hardware with these rather than the pooperhub.

Peter_
30-04-2011, 11:38
The Pooper Hub will do 400Mbit/s.
The VMNG300 will do 200Mbit/s.
The VM256 will do 50Mbit/s.

In short it's a non-issue. The hardware is well ahead of the tiers being offered and will continue to be so.
I thought that the 256 could only go to 38Mb.

Ignitionnet
30-04-2011, 12:02
It is a hybrid modem so can sync to 256QAM EuroDOCSIS channels.

The DOCSIS only modems on the ex-ntl side were the ntl-100, Terayon, etc, only Telewest provided them anything like recently as they ran DOCSIS only network until post-merger.

If a modem is connected to a BSR or 10k it is almost certainly handling 50Mbit on the front end (still the very odd 64QAM EuroDOCSIS channel around where downstream laser loading is high and they haven't been upgraded), actual throughput they can deliver to the devices behind them will vary.

TJS
30-04-2011, 12:10
Also; what is the difference between QPSK for upstream; and 16QAM for upstream?

---------- Post added at 11:10 ---------- Previous post was at 11:09 ----------

Oh, and ATDMA

Chrysalis
30-04-2011, 12:46
QPSK is a slower mode that has more tolerance for noise, so it gets used when there is notable noise affecting the port. It is around half of the bandwidth capacity of QAM16.

ATDMA vs DTMA shows you on docsis2 vs docsis1.

Skie
30-04-2011, 13:29
The Pooper Hub will do 400Mbit/s.
The VMNG300 will do 200Mbit/s.
The VM256 will do 50Mbit/s.

In short it's a non-issue. The hardware is well ahead of the tiers being offered and will continue to be so.

Actually the poopyhub has a chipset capable of 320Mbit/s. As it is also being used for the router side rather than purely the modem then the max is probably a bit lower than that.

http://www.broadcom.com/products/Cable/Cable-Modem-Solutions/BCM3380

Still, 300meg is a while away.

Ignitionnet
30-04-2011, 13:45
Actually the poopyhub has a chipset capable of 320Mbit/s. As it is also being used for the router side rather than purely the modem then the max is probably a bit lower than that.

http://www.broadcom.com/products/Cable/Cable-Modem-Solutions/BCM3380

Still, 300meg is a while away.

Hehe. That webpage is contradictory.

With total capability of over 500 kpps, the BCM3380 enables line speed rates, maximizes the eight downstream and four upstream channels of DOCSIS/EuroDOCSIS 3.0, with plenty of processing power to handle gateway and other advanced applications such as WiFi or VoIP.

Broadcom's highest speed DOCSIS/EuroDOCSIS 3.0 solution available with downstream up to 320 Mbps and upstream up to 160 Mbps enables competitive MSO offering

They've given the DOCSIS performance figures, 8 x DOCSIS downstreams are very approximately 320Mbps. EuroDOCSIS gets another 80Mbps out of the 64MHz.

The chip set bonds 8 channels and can handle 256QAM EuroDOCSIS so at very least the modem must be able to process 400Mbit/s else it'll start dropping incoming MPEG frames.

Chrysalis
30-04-2011, 13:55
yeah ignition there is no way the superhub will do 400mbit of internet throughput and especially with SPI enabled. Typically processing internet traffic is more cpu consuming than lan traffic and it can barely handle 400mbit of lan throughput. It supports 8 channels but that doesnt necessarily mean 400mbit as in theory 30mbit service can be supplied over 8 channels.

If I were to guess I would say 100-150mbit of internet traffic is its max capability with SPI on and about 200-250 with it off.

Ignitionnet
30-04-2011, 14:12
yeah ignition there is no way the superhub will do 400mbit of internet throughput and especially with SPI enabled. Typically processing internet traffic is more cpu consuming than lan traffic and it can barely handle 400mbit of lan throughput. It supports 8 channels but that doesnt necessarily mean 400mbit as in theory 30mbit service can be supplied over 8 channels.

If I were to guess I would say 100-150mbit of internet traffic is its max capability with SPI on and about 200-250 with it off.

Yes it does mean 400Mbit else it's not a certifiable chipset.

Broadcom claim the chipset can handle over 500kpps - this would give it the ability, before any kind of processing of anything else - to shift 5.8Gbps. They also note:

With total capability of over 500 kpps, the BCM3380 enables line speed rates, maximizes the eight downstream and four upstream channels of DOCSIS/EuroDOCSIS 3.0, with plenty of processing power to handle gateway and other advanced applications such as WiFi or VoIP.

As a reminder on the LAN side there is virtually no CPU involvement in LAN traffic, if it struggled with 400Mbps switched throughput it wouldn't be able to get near 100Mbps of routed throughput let alone the 200Mbps and more it's been up to in trials.

The firmware may be a PoS but I've no reason to doubt the capability of the hardware unless Broadcom are lying about it benchmarking at >1000DMIPS.

I reckon it can do 393.421Mbps with SPI enabled so long as the firmware is efficient.

See I can pull baseless figures out of my head too :)

Chrysalis
30-04-2011, 14:24
mine arent baseless tho and you know that.

before ssh got disabled I monitored cpu usage on it during various tests and we know the superhub self reboots during overload on speeds of over 300mbit.

packets per second and throughput are 2 different measurements, hardware can be capable of high pps but not necessarily high throughput. Likewise it can be capable of high throughput but not high pps.

for your information when I shifted just 70mbit of throughput over my lan the cpu usage was over 20%. I dont have cpu figures for higher throughputs as by then I lost ssh access but there is no doubt the superhub struggles to process high speed traffic flows.

Or do you have a superhub and real usage data? maybe I am mistaken and you not just relying on tech sheets right?

TJS
30-04-2011, 14:30
mine arent baseless tho and you know that.

before ssh got disabled I monitored cpu usage on it during various tests and we know the superhub self reboots during overload on speeds of over 300mbit.

packets per second and throughput are 2 different measurements, hardware can be capable of high pps but not necessarily high throughput. Likewise it can be capable of high throughput but not high pps.

for your information when I shifted just 70mbit of throughput over my lan the cpu usage was over 20%. I dont have cpu figures for higher throughputs as by then I lost ssh access but there is no doubt the superhub struggles to process high speed traffic flows.

Or do you have a superhub and real usage data? maybe I am mistaken and you not just relying on tech sheets right?

If you could ssh into it, does that mean its a linux based router? also how did you check the cpu usage? did you use top?

Chrysalis
30-04-2011, 14:31
incidently what is the lan side chipset? or is that broadcom handling both wan and lan.

Ignitionnet
30-04-2011, 14:33
Sorry but this statement:

Likewise it can be capable of high throughput but not high pps.

On its own shows you have no idea what you are talking about.

Given that data is delivered in packets in the case of most of our home networks no larger than 1500 bytes how exactly do you propose a device can have a high throughput without being able to shift enough packets to support that throughput?

It is the processing of these packets, via reading and rewriting the headers, updating state tables, etc, that is CPU intensive. The size of the packet itself is largely irrelevant with the exception that small packets tend to reduce maximum throughput because the overhead shifting a 100 byte packet is nearly identical to that of shifting a 1500 byte packet.

Given that you have no idea what you're talking about beyond mooching at an SSH shell on a device with broken firmware you'll have to forgive me if I don't take as gospel your opinion on what the hardware itself is capable of.

---------- Post added at 13:33 ---------- Previous post was at 13:32 ----------

incidently what is the lan side chipset? or is that broadcom handling both wan and lan.

You tell me? You apparently are au fait enough with the hardware that you can confidently predict its capacity.

Chrysalis
30-04-2011, 14:35
If you could ssh into it, does that mean its a linux based router? also how did you check the cpu usage? did you use top?

its linux based yes and top worked yes.

sshd is still running but now is blocked via firewall rules that are always enabled to block the port regardless of setting in gui.

I accept the limitations could be all down to bad firmware, however that is not proven and only a possibility and given VM's track record so far of fixing firmware issues I think the chances of them been successful in gaining a large performance boost from a firmware update is very remote.

A reason for the lan traffic been expensive to process on the superhub could in theory be that the firewall is not excempting it from inspection which would increase cpu utilisation but that is only theory but what is fact is the figures I got when testing.

TJS
30-04-2011, 14:36
its linux based yes and top worked yes.

sshd is still running but now is blocked via firewall rules that are always enabled to block the port regardless of setting in gui.

I accept the limitations could be all down to bad firmware, however that is not proven and only a possibility and given VM's track record so far of fixing firmware issues I think the chances of them been successful in gaining a large performance boost from a firmware update is very remote.

A reason for the lan traffic been expensive to process on the superhub could in theory be that the firewall is not excempting it from inspection which would increase cpu utilisation but that is only theory but what is fact is the figures I got when testing.

What kind of load averages did it show? and how much RAM does the superhub have e.t.c?

Chrysalis
30-04-2011, 14:37
ignition I have witnessed first hand hardware that maxes out cpu at high throughput using large packets with a lower pps than lower speeds at higher pps, it happens, I am not claiming to know why but do claim to know it happens. Thats why I am saying capability of high pps is by no means a guarantuee it can handle high throughput. Typically yes high pps is the cpu killer which is why many ddos attacks concentrate on that rather than bandwidth saturation now days but to claim every single bit of hardware will always handle high throughput if it has a high pps rating is wrong.

Ignitionnet
30-04-2011, 14:38
If you could ssh into it, does that mean its a linux based router? also how did you check the cpu usage? did you use top?

It's not running Linux.

See this thread. (http://community.virginmedia.com/t5/Up-to-50Mb-broadband/Interesting-stuff-when-ssh-d-into-superhub/td-p/379669)

Chrysalis
30-04-2011, 14:40
What kind of load averages did it show? and how much RAM does the superhub have e.t.c?

I cant remember off hand and I really should have saved the data but I wasnt expecting ssh to be shutdown so quick, I was only able to use it for one day. However i can do some tests on my spare superhub which still has sshd open.

Ignitionnet
30-04-2011, 14:42
ignition I have witnessed first hand hardware that maxes out cpu at high throughput using large packets with a lower pps than lower speeds at higher pps, it happens, I am not claiming to know why but do claim to know it happens. Thats why I am saying capability of high pps is by no means a guarantuee it can handle high throughput. Typically yes high pps is the cpu killer which is why many ddos attacks concentrate on that rather than bandwidth saturation now days but to claim every single bit of hardware will always handle high throughput if it has a high pps rating is wrong.

Care to show me some of this hardware that's maxed out.

If you could restrict it to networking equipment as opposed to PCs which will have an application layer component that would be a bonus.

Yes PPS count being high is an assurance of high throughput. That's why the specifications on routers show said count rather prominently and why within the specifications on their hardware they give throughput at certain packet sizes - because the constraint is the amount of packets that the various routing and switching elements can route/switch, routers couldn't care less about payload they just shove it onto a buffer. Even devices doing NAT still only concern themselves with headers, they just perform shallow packet inspection and read / rewrite TCP headers in addition to IP headers.

Or are Cisco, Juniper, et al are also wrong?

Sorry and all that but given you are trying to rewrite some fundamental facts about routers and switches you'll have to do a little better than 'ignition I have witnessed first hand hardware that maxes out cpu at high throughput using large packets with a lower pps than lower speeds at higher pps'.

Chrysalis
30-04-2011, 14:48
It's not running Linux.

See this thread. (http://community.virginmedia.com/t5/Up-to-50Mb-broadband/Interesting-stuff-when-ssh-d-into-superhub/td-p/379669)

what OS is it then?

looked like linux to me, wasnt bsd and wasnt solaris.

---------- Post added at 13:48 ---------- Previous post was at 13:42 ----------

Care to show me some of this hardware that's maxed out.

If you could restrict it to networking equipment as opposed to PCs which will have an application layer component that would be a bonus.

Yes PPS count being high is an assurance of high throughput. That's why the specifications on routers show said count rather prominently and why within the specifications on their hardware they give throughput at certain packet sizes - because the constraint is the amount of packets that the various routing and switching elements can route/switch, routers couldn't care less about payload they just shove it onto a buffer.

Or are Cisco, Juniper, et al are also wrong?

Sorry and all that but given you are trying to rewrite some fundamental facts about routers and switches you'll have to do a little better than 'ignition I have witnessed first hand hardware that maxes out cpu at high throughput using large packets with a lower pps than lower speeds at higher pps'.

If you really interested to get you exact model numbers I will have to check with the guy who was configuring the equipment. Back in the days when I was dealing with 'some' heavy DDOS we had to employ some new juniper routers which were brilliant at dealing with ddos due to a very high pps rating, but then we found issues with sustaining high throughput and getting cpu saturation even tho it had a pps rating way higher than the previous equipment. Consequently it got murdered on bandwidth saturation DDOS but was good at small packets high pps DDOS, normal traffic was also enough to saturate the cpu at high throughput.

If I were to guess I could say it could possibly be an issue with ram access from the cpu, if it has a poor cpu-ram bus then I would guess bigger packets which need more memory to process would hit the cpu particurly hard. I have no idea if a new firmware resolved it as we ended up changing the hardware. I also have found it not uncommon for hardware to max out below what its rated at in both pps and throughput. This is with both cisco and juniper.

TJS
30-04-2011, 14:48
What is considered high PPS, because when i download, or do a speedtest, activity monitor on my mac shows arround 3,000 pps

https://www.cableforum.co.uk/images/local/2011/04/3.png

Ignitionnet
30-04-2011, 14:52
what OS is it then?

looked like linux to me, wasnt bsd and wasnt solaris.

Netgear's own.

*Nix looks like this:


root@Wintermute:~# cd /
root@Wintermute:/# ls
bin etc lib mnt proc sys usr www
dev jffs mmc opt sbin tmp var

Chrysalis
30-04-2011, 14:58
I will take a step back and acknowledge this could well be just a firmware issue for the superhub issues, but if it is I do find it unlikely it will get resolved. In addition my past experience of poor performance from high pps routers could have been down to firmware.I will let you know when I have the model numbers of the routers in question as this was a good 6 years ago now when this happened, since then I have moved away from using own routers and simply rent capacity from DC's instead via dedicated servers. (far less hassle).

Ignitionnet
30-04-2011, 15:04
what OS is it then?

looked like linux to me, wasnt bsd and wasnt solaris.

---------- Post added at 13:48 ---------- Previous post was at 13:42 ----------



If you really interested to get you exact model numbers I will have to check with the guy who was configuring the equipment. Back in the days when I was dealing with 'some' heavy DDOS we had to employ some new juniper routers which were brilliant at dealing with ddos due to a very high pps rating, but then we found issues with sustaining high throughput and getting cpu saturation even tho it had a pps rating way higher than the previous equipment. Consequently it got murdered on bandwidth saturation DDOS but was good at small packets high pps DDOS, normal traffic was also enough to saturate the cpu at high throughput.

If I were to guess I could say it could possibly be an issue with ram access from the cpu, if it has a poor cpu-ram bus then I would guess bigger packets which need more memory to process would hit the cpu particurly hard. I have no idea if a new firmware resolved it as we ended up changing the hardware. I also have found it not uncommon for hardware to max out below what its rated at in both pps and throughput. This is with both cisco and juniper.

Why exactly would 10Mbps delivered within the least possible number of packets on a 1500 MTU link (854) require more memory than 10Mbps delivered within say 8,540 packets?

It would actually require less as the device has less headers to look through. Assuming it's a device with no ASIC or FPGA support there will be less traffic between CPU and rest of device with larger frame sizes.

I would check with the guy who was configuring the stuff and see if either he wasn't doing some odd configuration or there wasn't a software bug present.

I'm fine with the idea that there can be other bottlenecks. I'm not so fine with your claim that PPS doesn't indicate the routing capacity of the device's CPU.

---------- Post added at 14:04 ---------- Previous post was at 14:02 ----------

I will take a step back and acknowledge this could well be just a firmware issue for the superhub issues, but if it is I do find it unlikely it will get resolved. In addition my past experience of poor performance from high pps routers could have been down to firmware.I will let you know when I have the model numbers of the routers in question as this was a good 6 years ago now when this happened, since then I have moved away from using own routers and simply rent capacity from DC's instead via dedicated servers. (far less hassle).

Yes please.

If your guys were using the Junipers to do layer 7 inspection of traffic this would also explain the issues. The numbers quoted for their performance are usually not including whatever other functionality that is dependent on CPU there is.

With hard core commercial grade routers PPS is a function of switching fabrics and ASICs or FPGAs not CPU. CPU getting thrashed will usually indicate that a non-ASIC/FPGA assisted function such as DPI is being done.