Recent peak-time speed problems (Nov 08) - Please read this!

[#1]

I'm glad I'm not the only one that finds that even DNS lookups are noticeably slower when the network is under load (evenings etc)
I wonder if its just sheer processing load as they have to prioritise every packet that is returned to the user, if this is the case then pro users are likely to be equally affected as ALL traffic passes through the boxes according to what I've read on here and indeed the diagrams on Arbor Networks' own site
Now as I understand it the Juniper E-series routers also have traffic identification and metering capability so why not re-route pro traffic round the traffic management platform (I appreciate this maybe more easier said than done on a physical level) customers that use beyond 16GB of bandwidth can have their connections throttled but this would remove some processing load from the Arbor boxes surely and reduce gold drops?

[#2]

The Ellacoyas are very unlikely to be at capacity.
I'll post some stats from my Pro account when I get home.
B.

[#3]

Question.
Why do so called drops occur?
Is a drop defined as where the TTL value on the packet is exceeded because its been queued too long or is the hardware simply dumping traffic?

[#4]

1.  Your product type
PlusNet Broadband Premier Option 1
2.  Your current usage from the VMBU
Peak usage: 1.37GB
Off-peak usage: 90.14MB
Total usage: 1.46GB
3.  Your IP Profile, as reported by Plusnet
Your current broadband speed is 8000 Kbps
4.  Information from the BT Speedtester
Unable to test as always busy.
5.  A Speedtest from an external testing site.
Typical 'bad' test: as slow as dial-up!
'Good' test:
6.  Your gateway
You are currently connected to the "Telehouse North, LLU" gateway. This gateway is sometimes known as thn-ag2.
7.  Your exchange status
Results: Kemptown
Code: SDKMPTW
VP capacity at this exchange is currently showing as Green.
8.  Any other information
Speed is often slow in the evenings, from 5pm to midnight.

[#5]

you cant do the BT speed test as you are on LLU

[#6]

Hi
Just saw this post but was on AG02 and AG01 and seeing laggy web page downloads.




My IP profile is 8000kbps and useage just over 3GB this period. I normally get 5Mbps on average so this isn't as good as PN normally is.

[#7]

Quote from: techguy

Question. Why do so called drops occur? Is a drop defined as where the TTL value on the packet is exceeded because its been queued too long or is the hardware simply dumping traffic?

As far as I understand it the routers apply QoS and start selectively dropping packets when the network throughput increases above a certain level.  The type of packets dropped (and perhaps number?) increases as the traffic builds up.  Time of day may also be a parameter that goes into the equation.  The routers are intelligent enough to drop different amount of packets according to the type of customer account (by flagging packets accordingly).  Someone from PN who knows how the Ellacoyas are configured can hopefully explain this better.
To add to my previous question - if the lag is a DNS issue rather than throttling, is anyone running a local DNS resolver and if so is there a noticeable difference?

[#8]

Thanks for the response Mick
Yes it would make sense for some packets to be dropped to enforce a speed control.
Lets hope they sort the classification issue out though it could become a pain.
Not holding my breath for a response from PN for a response on this though as I think about it because it relates to network management which I suppose could be classed as commercially sensitive.
I'll see how it goes when I start really using my connection in earnest next year, hopefully they'll have sorted it out but if not I'll leave the line with PN and look at the BB market again.
PN are good to deal with but hopefully they will flush out the gremlins in the traffic platform.

[#9]

Quote from: techguy

PN are good to deal with but hopefully they will flush out the gremlins in the traffic platform.

This is not a new problem by any means, I think PN have had the Ellanoya installed for nearly 3 years, it seems they need some TLC at least once a year. The problems are the same. it does not fulfill the PN promise of 'no gold traffic' being dropped, the system is not as intelligent nor automagic as it was perceived to be.
SW.

[#10]

 
Hope it stays like this later on if I get back to check. 

[#11]

Companies such as Arbor Networks (who acquired Ellacoya networks) use the word intelligent in their sales literature which is pure nonesense and of course appeal to the accountants (this device will save you money)
To work properly the hardware needs to be executing the code properly or something is occuring as a result of increased traffic once a year, if this is the case then it needs to be raised with Arbor who are the ones in the position to implement a fix.
In all fairness, the poor soults that administer the platform are probably not the ones that made the decision to buy it (I know that is the case where I work where we've had systems that don't work as we like but we're stuck with them because a manager who has no contact with the system made a devision to buy it and ignores our complaints)
I'm sure the members of staff are doing their best to sort out the problems which are not of their making, hats off to them for that.
Interestingly, PN are used as a case study as a successful deployment, hopefully the issues are being raised with Arbor and a permanent fix maybe implemented some day

[#12]

Quote from: techguy

Interestingly, PN are used as a case study as a successful deployment, hopefully the issues are being raised with Arbor and a permanent fix maybe implemented some day

We're used as a case study because we have done much more with the Arbor Ellacoya solution than other providers have. We've not simply stuck their solution in and left it at its default settings, its completely integrated into Workplace (OSS), our Juniper E-Series Routers and probably has the largest customer specific configuration on it. 
I have not read this whole thread so not sure where the assumption (if it is one) that a fix is required from Arbor Ellacoya, this is simply not the case.

Quote from: godsell4

This is not a new problem by any means, I think PN have had the Ellanoya installed for nearly 3 years, it seems they need some TLC at least once a year.

Yes quite a while now we;ve had Ellacoya in our network and the TLC needed is just housekeeping, its a platform that when installed and integrated as much as ours is with other systems etc. we cannot just leave them alone to do their magic, the whole platform (talking traffic control here) needs regular housekeeping and this activity has dropped off but I'm making sure its picked back up, hopefully you can see the improvements of this already. 
I think godsell4 knows me from previous TLC moments, I won;t stop till the platform is back to the acceptable levels we advertise and I'll will be responsible for making sure housekeeping is maintained going forward (some more responsibilities I have won 🙂 )
HTH

[#13]

Hi _CN_
Thanks for the reply, I've no doubt that you will work to keep on top of it (To be fair I've intended to watch the presentation by PN's CTO about your deplyment as I find networking stuff fascinating and am looking to move into this area but have other things to take care of) so apologies if my comments are a little out of turn.
Best Regards
Mitch

[#14]

Hi _CN_,
Just as I thought that your intervention would be all that is needed to sort things out, it looks rather poor again both yesterday and tonight:

Date	            Download Speed	        Upload Speed	
Today 19:32	    3236 kbps (405kB/s)	357 kbps (44.6kB/s)
Yesterday 19:13	5010 kbps (626kB/s)	364 kbps (45.5kB/s)
Sunday 16:47	   6229 kbps (779kB/s)	372 kbps (46.5kB/s)
Sunday 00:40	   6185 kbps (773kB/s)	373 kbps (46.6kB/s)

Would moving from BBYW Option 2 to Pro fix this problem?

[#15]

Mick, has your line profile stayed the same?