dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
5269

odog
Minister of internet doohickies
Premium Member
join:2001-08-05
Atlanta, GA
Nokia BGW320-505

odog to xymox1

Premium Member

to xymox1

Re: Cox speed upgrade degraded my service

Feel free to send your modem MAC. I will however warn you in advance that you're not doing very scientific tests. You're testing latency to the gateway.... over a connection you're using.

Also increasing the speed of a connection doesn't change the CMTS configuration. Those are just bits in the cm config file telling it to be 150,000,000 instead of 50,000,000.

Do you understand how unbelievably catastrophic 50ms gateway latency would be? We would have dogs and cats living together, and mass hysteria. Every gamer in the entire PHX metro area would be camping outside our offices.

I'd also suggest thumbnailing the images, otherwise it makes this thread hard to read.

I'm not sure if this has been suggested but bring it back to the most simple config possible. Try connecting one computer directly to the modem, and don't use the connection AT ALL. I can also setup concurrent pings to both your modem and CPE and watch them for an hour.
nickphx
join:2009-10-29
Phoenix, AZ

nickphx to xymox1

Member

to xymox1
Again, no proof. Just rants.

You're expecting commercial service level agreements on a consumer level service.
You're nuts.
You're ranting about 2ms in gain.
You're choosing to say it's 15-20% increase in an attempt to make it appear worse than it is.

You're using ICMP to test.
This is not a reliable method to test or troubleshoot. It has already been stated countless times by COX people on this forum that ICMP is de-prioritized on their gateways/routers/equipment.

You're testing from a connection you're actively using. I wonder if that impacts testing?

You're not really showing any real problems, just ranting using an unreliable source of data for the reasons behind your rant.

Yes, I'm sure it's all a guise for COX to install content and protocol filters. They're also watching you pee. This is why I've lined my computer room with tin foil and chicken wire and maintain an AES256 encrypted vpn connection with multiple servers. They'll never be able to see which kitten videos my wife chose to watch today!

bbeesley
join:2003-08-07
Richardson, TX

bbeesley to odog

Member

to odog
said by odog:

I will however warn you in advance that you're not doing very scientific tests. You're testing latency to the gateway.... over a connection you're using.

What odog is telling you is spot on here. You can't reliably test for latency and jitter as granular you are trying to do with the software you are using because the client PC running the software could be introducing errors in the report due to CPU load and you have no idea if the devices in the path are priority responding to ICMP.

If you want to test this reliably, you would need to get a test set or a NID and test point on either end of the connection that could do RFC2544 or Y.1731 performance reporting.

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

2 edits

xymox1

Premium Member

ODog, Sorry for the wide charts. I use a HD resolution so I did not notice the wide images causing the text to be unreadable. I kinda figured most people had close to HD resolutions these days, but I would imagine most computers at work would be lower rez, sorry.
quote:
Also increasing the speed of a connection doesn't change the CMTS configuration.
Of course. But changing the speeds I think also brought with it a change in gear. Somewhere between me and my gateway something changed.

Im going to address all the comments above, so these are not directed at you Odog

Just a reminder I have been running this same setup with Multiping for 5 YEARS with the same latency charts. 5 YEARS. With once per second pings to the gateway with basically zero jitter and extremely consistent latency. FOR YEARS.

Suddenly, and the moment they switched over to the new speeds ( and changed some topology elements or gear or something ), I noticed my my browsing had odd, but short pauses. I didnt even look at my multiping for like a week. I initially also did not notice the increased speeds. I have a wireline speed router and high performance network and a very fast computer so even small changes in the connection are visible. I did not notice the speed increase at first. What I noticed was the odd small hangs and it just seemed slower. It was like each element of a web page load was now landing kinda at a more stuttery pace. It was no longer a clean even load. The VERY FIRST THING I DID was plug my main computer directly into the modem. Browsed some web pages and seemed to notice the same thing. It was odd.

That was the moment I ran speed test and noticed the speed increase. I also went over and looked at my multiping. The charts confuzed me as I had better speeds, but worse latency and it jittered and had swings. They looked just like when a section of the cable modem system has too many modems and during peaks it gets overloaded and latency and latency spikes go up.

As I mentioned before, multiping runs on its own computer. Its a stripped down XP with lots of services stopped there is literally nothing else running on the machine and all non essential service are stopped. It runs with 34 processes running and before starting Multiping it sits at 97MB of ram used. Its a extremely lean OS setup. Then the hardware is fairly extreme. ASUS Rampage III Xtreme MB, high performance ram, Adaptec RAID 0 SSD's, high performance NIC..

The performance of Multiping is fairly unaffected by the machine normally even on a reasonable loaded machine. I suggest you read up on the software as thats a important aspect of their design.

So running it on a high performance machine, that does nothing else, that does not even have a monitor or keyboard and just sits in a rack doing nothing is a fairly clean way to do this.

Yes I did plug it directly into the modem for a day and saw exactly the same latency spikes and swings.

I have charts of it pinging my router. It shows sub millisecond ping times and zero jitter. I have charts showing it pinging my modem thru the router, again less then a ms and zero jitter. Although you can see the router work as you can see ever so slight changes ( sub millisecond ) when the router gets loaded with traffic.

I have baseline charts showing my network up to my cable modem having extremely low latency and jitter as my baseline.
quote:
Do you understand how unbelievably catastrophic 50ms gateway latency would be?
Yes. Ive seen it.

If you look at my 30 min latency chart, you will see the base latency is 8-10ms. But there are these spiky short ( almost single packet ) jumps to 50. Kinda bursty. They are fairly rare like randomly a few every 30 mins and are related to peak internet usage times, with 20 and 30ms being more common. So the reason you dont have people going crazy is its a high jittery kinda thing, so its hard to spot at first.

In my case I saw it as periodic slight hangs in web page elements loading. Friends of mine who are gamers experienced it as very brief lag that came in short bursts at key times of the day that they had not experienced before 2/1/13

My experience and theirs matched what my charts showed.
quote:
I will however warn you in advance that you're not doing very scientific tests.
Well its way better then prob 99.99% of everyone else using Cox and I think its a reasonable way to look at things. It does have limitations, but in this case I think it has caught a real issue.

nickphx,, you obviously missed the chart that shows a jittery increase up to 50ms in spiky bursts. Im not talking about 2ms, im talking about 50ms. My 15-20% comment was as a overall average its actually 3-5 TIMES worse in spiky peaks. But your trolling so im gonna just ignore your whining and rants.

Bbeesley,
quote:
What odog is telling you is spot on here. You can't reliably test for latency and jitter as granular you are trying to do with the software you are using because the client PC running the software could be introducing errors in the report due to CPU load and you have no idea if the devices in the path are priority responding to ICMP.
completely valid cautions. How the gateway responds to ICMP is indeed important. I have charts of this going back 5 years and its been exactly the same and very very consistent. Also I noticed the 50ms random spikes in real use on port 80 web page browsing long before i even looked at the charting. The charting just confirmed, at least for me, what I was seeing.
quote:
If you want to test this reliably, you would need to get a test set or a NID and test point on either end of the connection that could do RFC2544 or Y.1731 performance reporting.
I agree completely. But im not going to go spend money to do that. I might be able to accomplish this in Linux and I have some tools on my Mikrotik router. However I would need the same thing on the other end of course.

I would think that Cox can do this. Maybe just pinging off my modem. It would have to be high rez and frequent. a single ping a min is not going to catch this.

I would love for it to be a easy to address problem. But I dont think it is.

odog, I will send you my info. Im sorry to cause issues, but im a tad upset that a speed increase i was looking forward to has been a let down for me. I do love the new speeds, I just hope that maybe the issue im bringing up can get settled down..
xymox1

1 edit

xymox1

Premium Member

I think im also seeing some actual packet loss during peak internet times.

Ive started a DSLreports Smokeping. Its pinging from california so its got a higher latency and the jittery swings might not be that visible, but packet loss would be.

»/r3/sm ··· 05fe.CA1
xymox1

xymox1

Premium Member

I do like the Y.156sam performance testing, in particular the Service performance test. Does Cox use this ? Is this published someplace for Cox ?

»en.wikipedia.org/wiki/Y.156sam

"Packet delay variation is a key performance indicator (KPI) for real time services such as VoIP and Internet protocol television (IPTV) and is not measured by the RFC 2544"

So if Cox is just focused on using RFC 2544 they might not be quite as aware of the issues I am seeing. However they must do some testing of this..
xymox1

xymox1

Premium Member

»www.youtube.com/watch?v= ··· XyzGl5oM
xymox1

xymox1

Premium Member

The obvious question is what are the SLA acceptable levels for Cox ? I doubt these are published. Where is pass/fail for each service ?

It would be awesome to be able to run EtherSAM on my connection for 24hrs doing a service level test.

ikyuaoki
join:2011-04-12
Wichita, KS

ikyuaoki

Member

Cox does not provide any SLA for residential services unless I'd direct point you to the cox business class services there, If you need the SLA.

chrisf8657
join:2002-01-27
Glendale, AZ

chrisf8657

Member

Exactly. Almost all residential services under the USA are best-effort and not guaranteed. I specifically remember a paragraph in the Cox TOS that states no speeds or anything is guaranteed.

ikyuaoki
join:2011-04-12
Wichita, KS

ikyuaoki

Member

said by chrisf8657:

Exactly. Almost all residential services under the USA are best-effort and not guaranteed. I specifically remember a paragraph in the Cox TOS that states no speeds or anything is guaranteed.

That is correct. any speeds or latency are not an gauranteed.

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

Obviously. However, my whole point of this thread was that my service dropped in quality when in fact it was supposed to be a upgrade..

50ms spiky latency is alot however.

There is NO doubt that Cox has its own standards for residential SLA and I was just wondering if anyone had any idea what these were.

So maybe there should be, or already is, a reasonable standard for the industry for residential service ? This sounds like a good thing to have based on Y.156sam

I think it would be good if there was some way to judge a residential ISP based on its numbers, not just "speed"...
xymox1

xymox1

Premium Member

Ive done some more testing and qualifying of my test methods..

I set Multiping to do 5 pings per second. I wanted to see a bit closer temporally what is going on.

I also added a ping at the same rate to my cable modem thru the router as a base line to show if any latency was creeping into my measurements from my testing computer, router or networking.

I see regular spikes in latency above 30ms that last anywhere from 2-3 packets out to a second. Some as delayed as 70ms after a quick glance thru the last few minutes.

The chart clearly should show that my testing rig does not seem to be introducing any amount of latency in its measurements.

This chart shows 60 seconds of pings..

xymox1

xymox1

Premium Member

Yea I am seeing packet loss too... the DSLReports Smokeping sees it too..

Its a lot like the long latency spikes, these are short duration.



The smokeping is showing them right now as a 2.83% average over 3 hours. As high as 98% loss..

»/r3/sm ··· 05fe.CA1

I looked back over the months leading up to the change and i had zero packet loss. So this is brand new and started at the moment I got changed over to the higher speeds..

2.8% packet loss over 3 hours is significant.
xymox1

xymox1

Premium Member

You know... This is bad enough that maybe it is just some issue that might get fixed. I was worried it was some issue that i would be stuck with forever, but I think this is bad enough it might indicate a actual issue that can be fixed..

Also i want to point out again, as I have earlier in this thread, I love my Cox service. I have for YEARS. Zero issues. My intereactions with cox reps has always been great. The Cox guys here on the forum have been awesome..

My *only* issue is that my upgrade was more of a downgrade. Hopefully we can fix the issue.

ikyuaoki
join:2011-04-12
Wichita, KS

ikyuaoki to xymox1

Member

to xymox1
Seriously, there are no SLA standard for residential service, you may want to considering that the business service if you do need the SLA what Cox business has offer.

there are some things that suggestion way to improve your connection if you are serious gamer then the follow some steps, you have to disable the QoS packet scheduler by uncheck the diaglog box and you'll have to disable the TCP nagle algorithm to improve the TCP ping times.

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

Your suggesting that I need to swith to Cox buisness at more then $1200 / mo because Cox "upgraded" and degraded my residential service ? Seriously ?

And hehehe, I remember doing those changes in the registery in XP like 10 years ago. But indeed very good advice for people who are new to windows tweaks.
thedunlap
join:2005-01-31
Naples, FL

thedunlap to xymox1

Member

to xymox1
I'm based out of Gville FL, the vast majority of my traffic gets routed through 56 Marietta (Atlanta). For the last 3-4 years, 20-30ms + increases in latency depending on the time of the day has been the norm, ever since QoS was implemented.

I feel your pain... but welcome to the Cox network.

Your worst day on Cox, is still generally better than your best on ATT or Centurylink .

If you want to get into a customer service discussion, how many techs understand the difference between upstream and downstream power levels?

In my anecdotal experience:
Comcast answer = 1/19 techs
Cox answer= 11/13

Techs skipping out of work early by stating that no one was home= Comcastic customer service (I've never once had this issue with Cox).

While far from perfect, in the last 7 years:

I've needed two node splits that required several months of communication; two neighborhood plant issues; and three wiring issues from poll to house.

In-general, its been better to be a Cox customer than not.

As an aside, Cox's billing department could use some serious retrofitting.

ikyuaoki
join:2011-04-12
Wichita, KS

ikyuaoki to xymox1

Member

to xymox1
That's right. what i am suggest. you may want to consider to make switch to the cox business class service that is SLA and if you'll able to shell out lots more.

EDIT: I have the two cable modems connected to my single PC that window 7 do the load balancing it is 136 dollars a month that has 600GB/cap month that is alternative to the costly of cox business class service.

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

As I have mentioned a number of times in this email, I have completly enjoyed my MANY years of Cox. This is mostly why im upset. It caught me completly be surprize. I was expecting even better performance when they upgraded my service. I did not expect there would be a tradeoff and latency spikes in the 50ms+ region. Ive just come to expect such great service this took me by surprize.

I have clients using Centurylink DSL.. Its, ummm,,, awful...

ikyuaoki, yea ive been tempeted to use 2 modems and load balancing. However I dont use the bandwidth. I dont think that would help my latency or jitter, i think that might increase my issues. One connection is great for my use. If only it was not so spiky in latency..

I think residential users can expect some QoS and minimum SLA. The cable co being able to get away with not having one needs to end. I think it time we have some standards that need to be enforced by gov regulation. A Cable co will drift to the most profitable rather then whats best for the client.. Lets face it, its not like you have a choice in cable companies. Monopolies need oversight.

I dont think I should have to spend $1000+ / mo just to get a SLA. Not to mention if I went buis it would not help my situation as its all connected to the same plant.

ikyuaoki
join:2011-04-12
Wichita, KS

ikyuaoki to xymox1

Member

to xymox1
Well, it's up to you to make a decide choice to use the two cable modems that is alternative to the costly of Cox business service or else you'll be going to be with the Cox business service that is SLA.
ikyuaoki

2 edits

ikyuaoki to xymox1

Member

to xymox1
Click for full size
speed, latency and jitter
additional details here what i have with this more detail about speed, latency and jitter on my two connection lines to my single PC.

my jitter is very good about 1ms to 5ms i gets.
Centerton
join:2009-06-13
Bentonville, AR

Centerton

Member

Can you do this on a video game console?

ikyuaoki
join:2011-04-12
Wichita, KS

ikyuaoki

Member

said by Centerton:

Can you do this on a video game console?

He could do this with the video game console be connected to the cable modem but the latency is not guaranteed on the residential service.

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

And I wonder what is happening with buis clients ? *IF* and thats a big *IF* this latency jitter is caused by some aspect of the "upgrade" then it might also be effecting buis clients.

BUT..

If they are using RFC 2544 for the standard testing model for the SLA then jitter is not part of RFC 2544 testing. Plus RFC 2544 is very limited in scope. Y.156sam is far more complete and a better basis for testing against a SLA.

The jitter im seeing might not be subject to the Cox SLA and so they would not need to address it. However Jitter MUST be part of a buis SLA...

OK.. I will go look up the Buis SLA if I can..
quote:
Can you do this on a video game console?
A video game console, and video games in general, require very little bandwidth ( Mb/s ) what they need is low and non-jittery latency. Its really ALL about latency with vid games. A lot of games can be played surprizingly well on DIAL UP surprizinly enough. So a gamer should be looking at latency and jitter. Lower is better.

Load balancing 2 modems requires a router that can do it. Its NOT trival. It also requires a bit of technical know how networking wise. Make sure you understand what all you need and what gear is required before you order 2 accounts from Cox. ALSO understand what load balancing actually does. You cant download 1 file twice as fast, well, unless your torrenting. But normally if its a single connection to something/some site it will not double your speed. »www.ezlan.net/loadbalance.html

With gamaing the way it sends single packets to one server it might not have any effect at all. Also if it gets packets from 2 IP's at once for the same person it might freak out. I would say Steam would be a likely causality. But I dont know that for sure, but I assume a online game server might get a bit weird if it saw gama data for one person coming for 2 IP's..

Just make sure your understand what load balancing is before you proceed.

Plus you need hardware to load balance. The Mikrotik routers can do it, cheaply. »wiki.mikrotik.com/wiki/L ··· alancing and the products.. »routerboard.com/ among others..

However load balancing is just one more thing to get in the way IMHO. I like simple and direct with as few a processing steps inbetween.
quote:
additional details here what i have with this more detail about speed, latency and jitter on my two connection lines to my single PC.

my jitter is very good about 1ms to 5ms i gets.
ikyuaoki, thats not really good performance for $130. Im paying $89/mo and get this. Its a bit low as its peak internet usage right now. At non peak times I get 175/35

xymox1

xymox1

Premium Member

I cant actually find any SLA's. I find a zillion references to them, but I cant find them to read...

I did find something impressive tho..

"Standards-Based. Cox has achieved certification from the Metro Ethernet Forum. This assures you that Cox Metro Ethernet services provide reliable, standards-based solutions that align with the industry’s direction for carrier Ethernet services."

So thats fully Y.156sam

Reading all about Cox Buis is impressive. No doubt. I am a buis and have a home office with a LLC I run from it. So I could do a buis account. However the pricing is so much higher then residential its really kinda outrageous. YES it comes a one hella awesome service, but oh man to get 175/35 would be truly stupid expensive just to browse the web and do some small work realted things.

The part I dont understand... How can they provide a different level of service over the same wire on the same DOCSIS channels from the same plant ? Is it provisioned differently ? so maybe my path once it comes from the DMTS is different ? My connection is still based on my wire all the back to the plant. That sure isnt going to change.

You know....

If they cant work out this jitter thing, maybe I will do a slower connection that is buis... But still. I will have to pay ALOT more..
xymox1

1 edit

xymox1

Premium Member

HMMM........

After more research....

I cant find anywhere that Cox Phoenix has a SLA for buis cable modem based service.

Ultra ( which i assume now has gone to 175/35 rather then the 50x5 they list ) is $369/mo with a 1 yr agreement. So its 3.5 times more expensive and I dont see mention of a SLA *ANYWHERE*..

»ww2.cox.com/business/ari ··· internet

I do see SLA's in other states, just not Phoenix Az..

ikyuaoki
join:2011-04-12
Wichita, KS

1 edit

ikyuaoki to xymox1

Member

to xymox1
Notice this: load balancing does not double your bandwidth, I configured the windows 7 via the registry editor to add the values to setup the load balancing via the randomize so the router hardware is not required to do a load balancing so however there are linux box does do load balancing for the two connections as well as windows 7 can do that to load balancing if configured proper setup via the registry editor settings.

load balancing is very good to easing the pressure of bandwidth pipe.

i did not say that my jitter 1ms to 5ms is bad.

secondly it is not guaranteed to have a stable latency unless you have to configured your PC to handle a good buffer performance in-housed computer instead of the cable modem buffer.

I configured my PC to have large buffer in the computer memory instead of the cable modem of limited small buffer queued hardware.

that all i am saying this jitter is very good at 1ms to 5ms unless if there's higher jitter values then that means i have a bad computer configuration as most likely be misconfiguration settings.

EDIT: here's the jitter status i gets

xymox1
Premium Member
join:2008-05-20
Phoenix, AZ
ARRIS SB8200
MikroTik CCR1036-8G-2S+

xymox1

Premium Member

i dont understand how you can change registery settings to get load balancing. How do you connect 2 modems to one network ? Each modem is going to need to do DHCP. So your running without a router then ? So no NAT and therefore one one computer connected to the modem(s). I just dont see how that could work at all. I *suppose* a computer could do dual NIC's with a modem on each nic. Then sure I *suppose* its possible for win do have some way to load balance.

But thats a very one computer solution. I have more then one device I need connected to my network and need NAT. So the correct way is to use a router that can be setup to do load balancing along with NAT and a firewall, DNS caching server and other things.

Yea I dont understand that at all. Can you provide a link on exactly what your doing. I was unable to find any references to this at all using google.

dvd536
as Mr. Pink as they come
Premium Member
join:2001-04-27
Phoenix, AZ

dvd536 to xymox1

Premium Member

to xymox1
how about WITHOUT the router in the mix!