Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[jabns]

Hi all,

Not been on for a while for various reasons but i should be on a bit more often again now.

My exchange(Fleetwood, LCFLW) is about to get enabled for BE/O2 and i was basically wondering how good/reliable they actually are. Also latency is a major issue for me as i host alot of game servers in my spare time.

I am currently paying £95/m for 3 three bonded MAX Premium lines and can no longer afford to keep it up and i am looking to lower it to about £60/m. Also as i am sure kitz is aware the fleetwood exchange is not performing to good in general at the moment and so i would like to move to LLU.

I just read rizzla's post about the DNS issues and was wondering if this had been resolved. If not can you recommend any good paid for DNS servers.

The email servers and web servers i am not to bothered about as i host my own.

As i said latency/routing is one of the most important issues for me as i do alot of gaming so any info would be great.

Thanks,
James

[kitz]

No... go away - theyre carp. *
Go Easynet/Sky/UKO



That bandwidth is mine all mine   


>> am sure kitz is aware the fleetwood exchange is not performing to good in general at the moment

Yep I'm aware
If its like this on Max Premium - I dread to think what it would be like on Max. 
Between still having a stuck bRAS profile and contention Im getting pretty peed off with IPStream


-----
* Just in case anyone didnt realise - I was joking.


----
PS I just hope leechnet doesnt bring the Be backhaul to its knees too.
You know what this exchange is like

[jabns]

That would be about 55mbit/s just for me in a tripple WAN load balancing mode  .

PS: I forgot to mention the 7.5 mbit upload  .

James runs....

EDIT: Now i think my USENET access would = INSANE LOL

I think i need to get started on this tripple WAN linux box. This is going to be a learning curve .

[kitz]

>> That would be about 55mbit/s just for me in a tripple WAN load balancing mode

You really really dont want to go Be   How about trying TalkTalk. 
My neighbour has a TT connection - they dont use that much - nice line stats - similar to mine
Excellent speeds...............






for a 512k connection

[jabns]

 

We can compete for bandwidth 

[kitz]

Ummm  ok then...

Maybe not TT - how about UKO?

[jabns]

Then i would have to have 4 lines to get the same speeds.

Looks like Be  .

Don't worry i only use about 200GB/m  .

[kitz]

Sorry James I took your subject OT

As regards to DNS - I was thinking perhaps OpenDNS  I already have them as backup DNS servers they seem ok.   I honestly dont know about any paid for ones since I normally only use ISP with OpenDNS.  Perhaps rizla would be best answering that.   mr_chris is away atm otherwise Im sure he would have contributed.   Any other thoughts on DNS are welcome...  but from what Ive seen chris's connection is fine.

I dont know how much Be will put in on the backhaul, but the thought of competing against a triple WAN load balancer just scared me to death.

[guest]

As regards to DNS - I was thinking perhaps OpenDNS  I already have them as backup DNS servers they seem ok.   I honestly dont know about any paid for ones since I normally only use ISP with OpenDNS.  Perhaps rizla would be best answering that.   mr_chris is away atm otherwise Im sure he would have contributed.   Any other thoughts on DNS are welcome...  but from what Ive seen chris's connection is fine.

I dont know how much Be will put in on the backhaul, but the thought of competing against a triple WAN load balancer just scared me to death.

I can't help with the current state of Be's DNS but I'm not aware of any radical improvement - just about everyone I know on Be uses OpenDNS. The problem with that is actually getting to the OpenDNS machines. I had ~5-10% packet loss on Level3 at peak times every day, in fact it got so that 5-10% PL was normal and what with DNS being UDP rather than TCP then a significant number of queries went unanswered.

The backhaul isn't like an IPStream backhaul. There will be a "master" exchange (not yours) which has the backhaul and several* "sub" exchanges fed into that. From my experience in Leicester things can go pear-shaped (in terms of congestion) VERY VERY fast indeed.

If you are in an area where Be are expanding into then you can also expect significant downtime as the network is built. For the first 2 months that I had Be you could virtually guarantee that the connection would be down for 12-18 hours every weekend - in fact Be inflicted more downtime on me in 2 months than BT did in 6 years. Oh and it won't look like your connection has gone down if you ping the first hop as they disconnect you at the aggregation point rather than at your exchange - no authentication and a static IP on Be remember so no RADIUS/DHCP servers to contact.

*my exchange was one of 7 "sub" exchanges for example

Edit - James you may wish to take a look at what Be do with multiple IP address assignments. You won't get them in a block like a /29. Oh no. You will just get 8 IP addresses randomly assigned to you. More of a "scatter" than a block really I took one look at it and thought "No thanks" but you might be OK with it.

[jabns]

Thanks kitz and rizla.

It s a shame that they are not making the most of the opertunaties they have in terms of their coverage.

With my current ISP i learnt the importance of transits and peers. Where ever my data is going i know that it will take the right route and i just don't know whether i can live without that.

What i might do is get 2 BE Pro lines and one UKO line and make a rule in the load balancer that all DNS queries go through UKO and then through BE if that is unavailable.

James

[kitz]

>> From my experience in Leicester things can go pear-shaped (in terms of congestion) VERY VERY fast indeed.

Hence concerns over the effects a few heavies could have..  then when you mentioned WAN loadbalancers why I went eeeeek.

I'm pretty certain that BTw perform some form of low level QoS on their DSLAMs - (I'm talking their newer models here - cause I know rizla was on one of the earlier BT enabled exchanges). 

But if you notice even when the exchange comes under heavy congestion, no-one complains about the latency. They specifically installed a pile of Juniper ERX's in 2003/2004 because of their ability "to distribute bandwidth more fairly" which wasnt happening on some of the older dslams.
ERX's have the ability to do the Bronze/Silver/Gold prioritisation and were the model that BT chose for the vast amount of exchange DSLAMs from 2003 onwards until they started using the Huawei MSANs.


>> in fact Be inflicted more downtime on me in 2 months than BT did in 6 years.


Again one of my concerns over LLU - I think theyve all been through it - Tiscali, TalkTalk/AOL etc.  From what I had observed Be seemed to be one of the better ones. 
Easynet do seem to have got it right, I dont ever recall seeing much complaints about their LLU. Although I have seen high packet loss and very probable signs of contention on the sky connections, I was told by UKO that sky and UKO have their own separate VPs.  This would appear to be true from observations elsewhere.

>> The backhaul isn't like an IPStream backhaul.

Indeed.  Some people think that BT give little away about their network, but information on the LLU Networks is even less.
I have spoken to Be in the past about other things, and they offered to give me a feed which is the same that Sam has.
Talking of which - the dates for Be around here:-

Code: [Select]

RFS date set: 30/04/2008 (originally 31/07/2007)


Blackpool Central 31/12/2006
North Shore 31/03/2007
Layton 31/3/07
Poulton 31/3/07
St Annes 31/03/2007
South Shore 31/07/2007
Cleveleys 31/07/07


Lytham RFS date set: 30/04/2008  (originally 31/05/2007)
Kirkham RFS date set: 31/03/2008
Weeton - Not available
Great Eccleston - Not available

[guest]

I'm pretty certain that BTw perform some form of low level QoS on their DSLAMs - (I'm talking their newer models here - cause I know rizla was on one of the earlier BT enabled exchanges).

Ah yes 2000ms+ latency with any traffic at all on the line and an ISP whose Technical Support manager stated that 200ms first-hop latency on an idle ADSL connection was normal. Happy days

[kitz]

>> Ah yes 2000ms+ latency with any traffic at all on the line

I remember seeing your l8nc graphs at the time
I believe it was the Cisco kit/dslams that BTw found werent able to handle QoS properly.  Ring any bells?

[guest]

It doesn't ring any bells but it doesn't surprise me. All I really remember re Cisco & BT was the CEF (Cisco Express Forwarding) business. As far as the DSLAM goes, it was an Alcatel but I can't remember any model details.

[jelv]

DNS: I've been using Treewalk for some time. It's your own local caching DNS that gets information from the authoritative servers.