Topic 'Re: What would cause this to happen' in forum 'Windstream' - dslreports.com

Re: What would cause this to happen

Sat, 08 Jun 2013 17:53:19 EDT

anon posted :

As you said, I want the lower latency for gaming. For some games, depending on server location, my latency is cut in half. It is a massive difference. My down speed is if little importance, I'd much rather have the lowest ping possible. I never noticed any instability when I was at 7ms, besides the usual congestion at 8pm-11pm.

I'm an FPS guy -- well at least many moons ago, I understand exactly what you're saying. The delicate balance here is incrementing CRC errors, rate adapt, and packetloss. CRC errors under FAST are retrasmits or loss, it's not like with INTL where you have parity and overlap to reconstruct the frame. FAST is great if you have a good line without line noise. If you're taking CRC errors under FAST then you're not going to be a happy FPS gamer depending on the quality of the netcode running and player-predict. You don't have a great line, you've got marginal SNR and aren't training at full rate and are likely taking CRC errors on the DSL side (line noise)

said by Higgins:

5/20/2013 - 7ms
Link up 6 US 763 DS 4632 (FAST:ADSL2)

Today
Link up 1 US 763 DS 5145 (INTL:ADSL2)

Looks like you're on a SpeedStream 4200 and/or 4300. I'm a fan of them. Modulation type is ADSL2, upstream isnt' fully trained to 768, and you're not fully training 6Mbit with over-provisioning.

said by Higgins:

My SNR is usually 7.5-8 which is bad. My Attenuation is 48-52. I have a technician coming Monday who will do absolutely nothing to fix the margins.

Yeah, you've clearly got line issues. I can see you're probably provisioned for 6Mbit but you're unable to train at that rate. If you don't care about downstream speed then I'd ask them to do a custom profile for you -- 3Mbit / 768K ADSL2 FAST -- 3Mbit/512K ADSL2 FAST or get to the bottom of your line issues. If you have static on the voice line report it as such, you may be on a bad pair. Too, the 5200/4200/4300's won't show upstream margins. I believe the TI AR7 in those modems (late rev 5200, prior was AR5) to be superior to the Sagemcom's.

said by Higgins:

I just called windstream for fun and asked if my connection was fast/intl. The person told me it was fast, and there was no way for me to tell what it is by looking at my logs...

Thanks for sharing your information, it's kind of sad I have to go to anonymous people on the internet for answers.

It varies, one of the primary drivers for my business line is the fact I can talk to Tier 2 and get escalated fairly quickly which is nice when I'm speaking to engineers and we can talk at the same level. If you asked "Am I INTL or FAST?" to a Tier 1 CSR you're going to get a BS response. If you say "I need you to pull my margins and tell me my modulation type and if I'm trained to the DSLAM as ADSL2 FAST or ADSL2 Interleave and I'd like to know the Interleave depth currently set on the DSLAM which is a (Cisco|Calix) fed by a DS3 in the CO)" you'll either get escalated or get a real answer... It makes no sense to put your engineers on the phone with grandma who says "My minnernets is running funny I can't play Pogo and my son in law in a computer guru he plays XBox online!" -- Tier 1 is there to handle the easy wins and escalate to domain experts as necessary.

Best of luck with your endeavours! I do not work for nor have ever worked for Windstream/Alltel.

Re: What would cause this to happen

Sat, 08 Jun 2013 16:50:07 EDT

Higgins posted : As you said, I want the lower latency for gaming. For some games, depending on server location, my latency is cut in half. It is a massive difference. My down speed is if little importance, I'd much rather have the lowest ping possible. I never noticed any instability when I was at 7ms, besides the usual congestion at 8pm-11pm.

I found the bit in the logs which identify if I'm on fast/intl:

5/20/2013 - 7ms
Link up 6 US 763 DS 4632 (FAST:ADSL2)

Today
Link up 1 US 763 DS 5145 (INTL:ADSL2)

My SNR is usually 7.5-8 which is bad. My Attenuation is 48-52. I have a technician coming Monday who will do absolutely nothing to fix the margins.

I just called windstream for fun and asked if my connection was fast/intl. The person told me it was fast, and there was no way for me to tell what it is by looking at my logs...

Thanks for sharing your information, it's kind of sad I have to go to anonymous people on the internet for answers.

Re: What would cause this to happen

Sat, 08 Jun 2013 16:36:37 EDT

anon posted :

said by Higgins:

If I understand it correctly the lower frame interleaving results in lower latency but a possible increase in the error/noise on the line. The higher frame interleaving is more resilient to error/noise at the cost of latency. Perhaps this could explain why my DSL stats were showing more Err/sec when my latency was 7.

Exactly!

said by Higgins:

What might I look for in the logs to determine if this is the case? I have my current logs when my ping is 35 and I also saved my past logs after my ping dropped to 7.

Usually, depending on modem, INTL or FAST will appear in the logs with various modulation types such as ADSL2+, G.DMT, ANSI, etc. In the past days of Alltel, I actually requested ANSI with FAST over G.DMT.

said by Higgins:

If this is the case, would it even be possible for windstream to change something on their end? The ISP controls the use of interleaving, but is this a function that can be switched on/off easily?

It is something they could do on the DSLAM, I've had it done before because I didn't want the latency overhead, but this was back in the Alltel days. The real question is why would you want to do this -- if you're chasing latency 35ms makes little difference aside from first-person shooters and high precision timing like NTP. It also comes at a cost of speed and stability.

Conversely, without INTL, you've already noted that you were taking increasing errors/second, of which I assume are CRC errors, not FECN. Consider OSI Layer 4 transport protocols, some are stateful, others are stateless. UDP which powers most VoIP, teleconferencing, gaming, content delivery, etc is stateless without error correction. Noise or CRC errors which are uncorrectable on the line will result in packetloss and drops, you'll never get that UDP datagram again unless your client software is handling state and error correction. TCP on the otherhand is stateful and you'll either lose a control flag or miss a PSH which requires the sender to PSH again as you've never ACKed it. You'll see a reduction in speed and network health/performance.

Honestly, I'd keep INTL, 35ms is nothing for the benefits you're getting from it. Enough errors on your line and the DSLAM/modem will start doing rate-adapt and start dropping your speeds until the incrementing errors cease. If you only want to do FAST then you need to chase down the source of the noise. INTL is great for handling line noise.

Share some margins, your logs, etc and I'd be happy to make suggestions. Wish you the best. There are some key areas that are clearly oversubscribed but others such as mine have been great after transport upgraded the CO to a DS3 instead of four T1s. Aside from my most recent ~18 H impact relating to the 3rd hop I've not had any issues with Windstream/Alltel for about 5 years and consistently see the bandwidth I pay for.

Re: What would cause this to happen

Sat, 08 Jun 2013 16:13:43 EDT

Higgins posted :

said by johnbor:

latency = ISP

ISP = WINDSTREAM

Conclusion
WINDSTREAM = TERRIBLE ISP PROVIDER

As with everyone who reads this forum, I have no other options. I feel like I have to do homework for possible solutions because they wont offer any.

Re: What would cause this to happen

Sat, 08 Jun 2013 16:06:02 EDT

johnbor posted : latency = ISP

ISP = WINDSTREAM

Conclusion
WINDSTREAM = TERRIBLE ISP PROVIDER

Re: What would cause this to happen

Sat, 08 Jun 2013 15:38:42 EDT

Higgins posted : I just read through the the interleaving and fastpath bit on wikipedia and that does seem like exactly what is happening. If I understand it correctly the lower frame interleaving results in lower latency but a possible increase in the error/noise on the line. The higher frame interleaving is more resilient to error/noise at the cost of latency. Perhaps this could explain why my DSL stats were showing more Err/sec when my latency was 7.

What might I look for in the logs to determine if this is the case? I have my current logs when my ping is 35 and I also saved my past logs after my ping dropped to 7.

If this is the case, would it even be possible for windstream to change something on their end? The ISP controls the use of interleaving, but is this a function that can be switched on/off easily?

Re: What would cause this to happen

Sat, 08 Jun 2013 14:28:59 EDT

anon posted :

said by Higgins:

Anyone educated on these matters that could provide some insight on the cause of something like this happening?

This latency reduction, especially when it corresponds to your modem reboot and is noticed on the first hop in a non RFC 1918 network (in your case 192.168.0.0/16), is most likely the difference between Interleave and Fast Path training to the DSLAM. Interleaving is used as a form of error correction on the DSLAM and modem and is described in more detail here »en.wikipedia.org/wiki/As ··· fastpath and »www.kitz.co.uk/adsl/inte ··· ving.htm

Your ~35ms is very close to what one would expect with an interleave depth of around 64 frame interleaving.

Hope this helped, if you're willing to share your logs from your modem training up to the DSLAM we can provide more information including your modulation type.

What would cause this to happen

Sat, 08 Jun 2013 13:22:01 EDT

Higgins posted : I am not educated on how internet routing works but something seems wrong here. My latency to the windstream hops was ~35ms for a long time. In January I called and the support I talked to said they would do some tests or something. Four hours later my latency to the windstream hops dropped to ~7ms, which resulted in a massive latency decrease to all sources. This lasted ~4 months. In May I noticed my latency was randomly back up to ~35ms to windstream, causing latency to go up across the board. I called support again and even made a thread on the support forums here. They said my line would be re provisioned. A week later my latency randomly dropped to ~7ms, they claimed nothing was done on their end. This lasted 17 days from 5/20/2013 to 6/6/2013. I went to bed at 12AM 6/6/2013 with ~7ms. I woke up 8 hours later back at ~35ms, where it remains today.

I have a screenshot of the ping drop happening in real time (to google.com). It occurred with a modem reset. I also have some graphics from a few months ago showing the hops at ~7ms.

I have no knowledge of how internet networking functions It's like I'm being routed to different servers on windstream.

Anyone educated on these matters that could provide some insight on the cause of something like this happening?