Superhub : Superhub & Virgin Media Business

[qasdfdsaq]

Buffering on the DOCSIS layer is usually configurable based on information sent from the CMTS to the modem (i.e. config file).

Has anyone actually done any tests to measure the upstream/downstream buffers on the SH? (e.g. Netalyzr or the few better around)

[craigj2k12]

I can test the superhub, but nothing to compare it to, if someone could run it on the VMNG, but then that would make it a bit unfair as its on a different part of the network

[qasdfdsaq]

The buffering is configured as a modem function so the part of the network shouldn't matter (unless they use different buffering setups on the same modem in different areas, which is possible, in theory)

[craigj2k12]

http://n1.netalyzr.icsi.berkeley.edu...9-402e-95d6/rd

[qasdfdsaq]

Network buffer measurements (?): Uplink 229 ms, Downlink is good +

Unusually low for VM. Still, Netalyzr is not one of the better buffering tests, I just can't dig out the good one I used recently

[Chrysalis]

Originally Posted by Sephiroth

Craigie

I suppose you're right about the amount of latency added in DOCSIS purely down to interleaving.

So it's back to that buffering tjing. Bonded channels complicate matters. Data doesn't arrive in sequence, and each channel can have different impairments; hence the high amount of buffering necessary to assemble a packet. Then there's the interleaving to further assist the FEC functions which adds to the buffer requirements.

So you can't avoid higher latency in DOCSIS and my sense tells me that it would increase further with 8 bonded downstream channels.

Or have I misunderstood something?

except the excess buffering is on the upstream and that currently has no bonding?

[Sephiroth]

Craigie

The main truth of my previous post was that bonding means that packets need to be re-constructed from the jumble, which includes jumble cause by interleaving. The higher the number of bonded channels, the greater the amount of re-assembly needed. Equals latency.

[craigj2k12]

Originally Posted by Sephiroth

Craigie

The main truth of my previous post was that bonding means that packets need to be re-constructed from the jumble, which includes jumble cause by interleaving. The higher the number of bonded channels, the greater the amount of re-assembly needed. Equals latency.

It will increase, but not by a noticeable amount, whereas bonding 8 channels compared to 4 will add <1ms, buffering more at the modem can mean 10ms+ (obviously depending on how many are buffered) - either way - the latency difference between superhub and VMNG isnt to do with channels, as when I was testing both, I only had 4 downstreams on both

And if you find Kymmy's TBB monitor, thats quite good (for a VM graph) and she is connected to 8 downstreams

[ccarmock]

Latest issues with the Business service.... The congestion causing poor performance has been improved greatly. Speeds of 48 Mb/s most of the time seem the norm again. Getting around 5.4 Mb/s up too which is good too.

However in Friday all day theer were drops and loss of sync. Suspected duff Superhub but then found to be due to a low SNR issue in the area. Supposedly fixed Saturday morning. Indeed the service was stable until Saturday evening and then overnight my think broadband graph shows lots of drops, then at 8 am dropped permanently with the blue light on solid on the Superhub. reboot fixes it.

Faults tell me they are seeing 50+ tunnel drops over the past 24 hours and I am likely to be suffering from a known firmware issue in the Business Superhub which causes a tunnel drop with UDP packets. I haven't experienced that with an IP Phone I have here, but do have a VPN tunnel setup too.

The person who diagnosed this as a firmware issue did say there are some limited field trials of a new version underway right now, so it may not be long before release.

The lady I am talking to in business complaints has arranged for me to be contacted by the team next week potentially to participate in that trial.

[itcomms]

don't expect the firmware fix to be bug free from previous experence.

I have put a Draytek 3200 with QOS control in on our connection, hoping to use this router when we go back to 50Mb

the fixes are taking way too long to be released. a fix should be sent out within a few days of the bug being noticed / reported.

[ccarmock]

Agreed - they should be able to turn these around far quicker. Godo to hear a new version is in test thoguh.

[APS]

Any news on this yet. Our service has been pretty good, although we just had an incident today where we lost connectivity even though the SH said it was connected. I guess it was an issue with the broken tunnel (a known issue) as a reboot fixed it. Better firmware would be good!

[ccarmock]

I have the beta version of the business superhub firmware on my box now. I've had it there just over a week. So far most of the issues appear to be fixed. The Microsoft Exchange problem with some inbound messages seems to ahve been fixed in that the MTU size of 1460 workaround can be removed and messages through ok.

Also the problem with the tunnel drops seems to have been fixed too. Also I get no more reboots of the Superhub.

An issue not yet fixed is the occasional lockup of the GUI when tryign to login to the Superhub.

I gather this version is looking good though and that it should turn into a full roll out shortly.

There was a major outage this morning - for me a reboot of the Superhub didn't bring service back - it seemed down from 4am aprox until around 9am, and came back without me doing anything.

[Nickjc]

I had the 100Mb residential service and Superhub installed yesterday and yes I am now unable to receive mail on my SMTP server. I note that the that the problem appears MTU related and we're waiting for new firmware. Unfortunately my residential Superhub does not appear to have the option to change the MTU setting to 1460 as suggested. Any ideas?

I've also tried the Superhub to modem mode with my old Netgear FVX538 router, however I get a very choppy connection and bandwidth chopping between 15Mb and 50Mb in this configuration. This config was working fine with VM's 50Mb modem.

Can anyone recommend a replacement modem/router that actually works with 100Mb service?

[Skie]

If you still have the 50meg modem get that connected and re-activated.