Where are my Ping Plots?

[Neffneff]

yep i got the same figure..1300. and mostly from the server itself, highest spike i have had from OT in the last 3 hours is 5-600ish...from what i can see (again no expert) opentransit is supplying us with its normal lag spikes all the time...but someting dogy is going on with the servers...compounding the problem.

 

[UndyingAngel]

yep i got the same figure..1300. and mostly from the server itself, highest spike i have had from OT in the last 3 hours is 5-600ish...from what i can see (again no expert) opentransit is supplying us with its normal lag spikes all the time...but someting dogy is going on with the servers...compounding the problem.

I agreee

 

[Neffneff]

make that peak number 1500ish :O

 

[Stasi]

cant care less, just fix it

 

[Chronictank]

3 hour sample for 5/12/2004 ending about 2am

6 hour sample ending about half 7

Those pretty red lines are what destroyed the ml4 raid 2 days running, hope they are of help

 

[Neffneff]

Dunno if it's a light at the end of the tunnel..but i logged onto daoc at someitme between 11pm and 12 GMT...its 2.5 hours later and i havnt seen a single red square or lag spike...

 

[RS|Phil]

I have a question -

How is it possible that everyone lags at the same moment server wide, or at least zone wide, in they're all from different parts of Europe?

If you're looking for a fault in transit surely the bottleneck must be very, very close to GOA HQ, not dispersed around the European Continental shelf?

Just a thought...

 

[RS|Phil]

Oh and do you still need any plots?

 

[Xalin]

Yes keep the pingplots coming please. Always room for more pingplots.

 

[Sendraks]

If you're looking for a fault in transit surely the bottleneck must be very, very close to GOA HQ, not dispersed around the European Continental shelf?

Have you not read this thread and not understsood what the purpose of the ping plots is?

 

[Gamah]

Weee I got the 100th reply to this thread

SEND YOUR PING PLOTS!

 

[RS|Phil]

My point is that at a glance it looks to be a common problem no matter what a person's locale, hence would it not be easier to trace the problem from the point of origin (the servers) to a destination in each of the client's countries?

As I said, it effects people all over the world at the same time so it's not something that's disseminated around the globe , it seems something more centralised - hence my wondering why the traces from all over to GOA, why not go from GOA to several places - thar'd just as easily identify the bottleneck, no?

 

[Grimmen]

My point is that at a glance it looks to be a common problem no matter what a person's locale, hence would it not be easier to trace the problem from the point of origin (the servers) to a destination in each of the client's countries?

As I said, it effects people all over the world at the same time so it's not something that's disseminated around the globe , it seems something more centralised - hence my wondering why the traces from all over to GOA, why not go from GOA to several places - thar'd just as easily identify the bottleneck, no?

Thats a good a point

 

[IainC]

My point is that at a glance it looks to be a common problem no matter what a person's locale, hence would it not be easier to trace the problem from the point of origin (the servers) to a destination in each of the client's countries?

As I said, it effects people all over the world at the same time so it's not something that's disseminated around the globe , it seems something more centralised - hence my wondering why the traces from all over to GOA, why not go from GOA to several places - thar'd just as easily identify the bottleneck, no?

Because the chances are that it's not a problem with a particular hub per se, it's more likely the routing between one hub and another. It's not as simple as saying 'Hub number 3 is broken, let's fix it', it's more like when hub 3 connects to hub 4, there's a problem. These connections are often different over time as OT switch bandwidth around to cover demand and won't necessarily be the same both ways - in other words connecting to London from Paris won't usually take you the same route as connecting to Paris from London. This is why pinging hubs around Europe from Goa HQ won't show the problem.

 

[RS|Phil]

That was also one thought but I dismissed it thinking it would be, essentially, one-way lag if that was happening. Ofc I aren't entirely sure how the client determines lag and latency, when causing link deads.

Anyway,I digress. This was certainly more informative than the usual FH thread, no?

 

[Boni]

Ill rephrase and ask my question again, making myself a bit clearer, anyone feel free to place 200plat on my chars and ill go away and pretend im just some clueless plastacine skull...

some facts?

a). It was very laggy this weekend, not just 'oh my PC is shite and I stutter a bit' or 'ooh im using a 33kmoden and a piece of string to connect', but the type of lag where Freddys House, alliance, group and say channels are constantly spammed with peopel saying 'FFS at this lag' and 'im logging/quitting/just gonna be rude at goa on FH cause this is no fun and im a vindictive teenager with no toy to play with'.

b). A lot of people have done pingplots for these periods, and while they all show a bit of lag around the world, most have one thing in common, large peaks and bad average performance from the last two stops.

c). As far as I can workout (and if this isnt the case please correct me) the last two stops are representative of the internal network/servers/routers, stuff that is not actually part of opentransit.

 

[RS|Phil]

 

[Tharion]

Ill rephrase and ask my question again, making myself a bit clearer, anyone feel free to place 200plat on my chars and ill go away and pretend im just some clueless plastacine skull...

some facts?

a). It was very laggy this weekend, not just 'oh my PC is shite and I stutter a bit' or 'ooh im using a 33kmoden and a piece of string to connect', but the type of lag where Freddys House, alliance, group and say channels are constantly spammed with peopel saying 'FFS at this lag' and 'im logging/quitting/just gonna be rude at goa on FH cause this is no fun and im a vindictive teenager with no toy to play with'.

b). A lot of people have done pingplots for these periods, and while they all show a bit of lag around the world, most have one thing in common, large peaks and bad average performance from the last two stops.

c). As far as I can workout (and if this isnt the case please correct me) the last two stops are representative of the internal network/servers/routers, stuff that is not actually part of opentransit.

Totally agree, funny thing is GOA acts like this started last week or so.
This has been going on for the last month and a half.

And all this GFX lag talk from these forum moderators, I think ppl can see the differance on a red F, and red P and L. And ppl lag no matter where in the realm you are. So rule out that crappy excuse!

 

[sibanac]

Ill rephrase and ask my question again, making myself a bit clearer, anyone feel free to place 200plat on my chars and ill go away and pretend im just some clueless plastacine skull...

some facts?

b). A lot of people have done pingplots for these periods, and while they all show a bit of lag around the world, most have one thing in common, large peaks and bad average performance from the last two stops.

c). As far as I can workout (and if this isnt the case please correct me) the last two stops are representative of the internal network/servers/routers, stuff that is not actually part of opentransit.

I think you need to look again at the plots, the problems are generaly in between london.opentransit.net and *.pastourelle.opentransit.net thats where the high latancy and packetloss starts, the rest is probably a cascade effect

 

[Boni]

I think you need to look again at the plots, the problems are generaly in between london.opentransit.net and *.pastourelle.opentransit.net thats where the high latancy and packetloss starts, the rest is probably a cascade effect

Ive looked several times at the plots, if you could examine say the two plots (thumbnails) I did over a few hours for 2 two very laggy periods this weekend youll see exactly what I mean The dont really point to *.pastourelle.opentransit.net etc at all, and I dont buy the cascade effect, as the lags I am pointing to are at the top of the waterfall so to speak...

 

[Boni]

^ edit, my thumbnails do show evidence of bad performance from OT, but they show worse performance from the last two stops, so if you where gonna fix the lag your best approach would be to start there.

 

[sibanac]

Ive looked several times at the plots, if you could examine say the two plots (thumbnails) I did over a few hours for 2 two very laggy periods this weekend youll see exactly what I mean The dont really point to *.pastourelle.opentransit.net etc at all, and I dont buy the cascade effect, as the lags I am pointing to are at the top of the waterfall so to speak...

on your first thumb you got over at hop 11 wanadooportalis.opentransit.net you got 67% loss, that means that anything after that hop aka the GOA network will be screwed aswell.

on the second
you are starting to get high latancy at hop 4 (ntl) and then some more buildup from hop 9 on.


and goa is the end of the waterfall, the stream goes from your pc to GOA.

If you look at some other plots you'll see that packet loss is always in the opentransit network. and while 400ms latancy is anoying it shouldnt cause ld's and a huge amount of lag, its the packetloss that is causing the grief

 

[Matchstick Man]

Unplayable on weekends, something needs to be done. They must have enough ping plots to prove there's a problem to OT now ???

 

[popa]

my pingplot

http://img.freddyshouse.com/forums/images/attach/png.gif
png.gif



here u have mine too

 

[Boni]

on your first ..

Sorry those are two plots from ages ago, if youll check the two in this thread, I trust them a lot cause they are both:-

a). Current to this weekends lag (so we dont get sidetracked talking about some lag I had 2 month ago).

b). Accurate, they are both 1-3hr plots, and ive even saved the .png in a format thats readable. I played the full time they where taken and experienced and recorded the lag.

I dont know about your 'cascade' idea, im not really an expert on how this software works, but I personaly trust it to provide information about each node. I would have also thought the the majority of traffic in the game comes from the servers to the user, only some minor UI data goes the other way. If your right about there being a cascade effect shown here, then all this pingplotting would be useless as its showing a). the wrong direction, b) corrupted 'cascaded' data.


here ill thumbnail them again for you, saturday and sunday this weekend.

 

[Matchstick Man]

You had it easy Think my broadband router slows me down a bit, but most of the week i get about 27 average.

 

[Boni]

Unplayable on weekends, something needs to be done. They must have enough ping plots to prove there's a problem to OT now ???

Now thats a nice plot, hours long, and phew look at them numbers on the last two hops... Im glad I went down teh pub that night.

 

[Darzil]

Aye, I'm finding the last two hops the worst, particularly recently.

When the problems first started (The infamous OpenTransit maintenance time) there tended to be many, many route changes (which aren't logged by pingplotter when you save). There were packet losses associated with this. I presume that a faulty router was leading to loss of connection, which made the working routers route around it. This problem seemed to get fixed, and there were a few days (for me) of much better play.

In the last few days (was away last week, so didn't check), the problem has seemed to be in the last two hops, sometimes the last three.

Darzil

 

[Darzil]

Because the chances are that it's not a problem with a particular hub per se, it's more likely the routing between one hub and another. It's not as simple as saying 'Hub number 3 is broken, let's fix it', it's more like when hub 3 connects to hub 4, there's a problem. These connections are often different over time as OT switch bandwidth around to cover demand and won't necessarily be the same both ways - in other words connecting to London from Paris won't usually take you the same route as connecting to Paris from London. This is why pinging hubs around Europe from Goa HQ won't show the problem.

To make matters even less clear, routers tend to automatically re-route traffic around faulty routers, making it harder to see what is going on. In fact this re-routing is often the cause of LD's, as when it happens, there are often a few confused packets of information around <grin>.

If networking was as simple as point A sends to point B sends to point C, it'd be easy to find problems. Unfortunately they are far more complicated than that. So complicated in fact that simulation programs are needed to try to predict the behaviour of networks. Mathematics alone isn't good enough.

It's also a bit frustrating that the tools to debug this sort of problem are scarily expensive. Packet sniffers are generally a few thousand quid, and even then, are only really useful if you know what you are looking for (hence the ping plots). I know I'd like such technology to debug communication issues at work, however I don't get the luxury.

To put in in perspective, one faulty network card or cable on the network that the last hops are on, sending out malformed packets, can choke that network to a crawl. I've seen it happen here. With a packet sniffer, you can pull out the malformed packets and see what their source IP address or MAC address is, and then remove/replace that device. Without one, you can pull network cables, wait a few minutes for the network to settle, and then see if the problem goes away. These problems take a while to build up and then settle down. That can take a long while, and wouldn't be something I'd want to try on a Server farm that is supposed to be up 24/7.

Darzil

 

[Xalin]

clicky