EPTRo16 Cast Delay Explanation - Page 3
| Forum Index > TSL2 Forum |
|
Fzero
United States1503 Posts
| ||
Macavenger
United States1132 Posts
 | ||
|
kiendudu
Netherlands273 Posts
| ||
|
R1CH
Netherlands10342 Posts
On January 31 2010 07:46 jalstar wrote: also why are these things not tested beforehand? feels weird to be a guinea pig It was tested as best I could, unfortunately the majority of the TL users already know how to watch streams and which software to use, so requesting them to test doesn't really help expose any problems. When we go live with something like the TSL, we have people trying to load it in all kinds of confusing ways such as their browser, winamp, realplayer, GOM, etc, combined with F5ing causing many reconnects. The server itself can easily handle thousands of connections so it isn't so much a capacity issue as some combination of people connecting, their video player reconnection behaviour, what initial headers are sent, etc causing some random bug. For the more technical among you, the server is coded in perl and makes use of libevent with the Event::Lib bindings. Event::Lib claims it has been tested in production use with MailChannels (a perl based email filter) with thousands of connections, so I thought it would be ok for the TSL. What happened is the program simply quit back to the bash prompt without even a single error or warning, even running with perl -w and with the Event::Lib exception handlers set up. To me this would indicate some kind of underlying problem possibly in Event::Lib or libevent itself rather than our server code, something I haven't been able to reproduce yet despite having thousands of connections and pushing a gigabit of traffic. There's just a limit to how much you can do in a test environment vs what the real world can throw at you. | ||
|
Grobyc
Canada18410 Posts
thanks for all your hard work, and no need to apologize, you're doing us all a favour | ||
|
ToN
Canada245 Posts
On January 31 2010 12:05 R1CH wrote: It was tested as best I could, unfortunately the majority of the TL users already know how to watch streams and which software to use, so requesting them to test doesn't really help expose any problems. When we go live with something like the TSL, we have people trying to load it in all kinds of confusing ways such as their browser, winamp, realplayer, GOM, etc, combined with F5ing causing many reconnects. The server itself can easily handle thousands of connections so it isn't so much a capacity issue as some combination of people connecting, their video player reconnection behaviour, what initial headers are sent, etc causing some random bug. For the more technical among you, the server is coded in perl and makes use of libevent with the Event::Lib bindings. Event::Lib claims it has been tested in production use with MailChannels (a perl based email filter) with thousands of connections, so I thought it would be ok for the TSL. What happened is the program simply quit back to the bash prompt without even a single error or warning, even running with perl -w and with the Event::Lib exception handlers set up. To me this would indicate some kind of underlying problem possibly in Event::Lib or libevent itself rather than our server code, something I haven't been able to reproduce yet despite having thousands of connections and pushing a gigabit of traffic. There's just a limit to how much you can do in a test environment vs what the real world can throw at you. dam dude, where did you learn all these things O.O. Or rather where could done even start?  I also would like to thank you for all the things you do. =) | ||
|
RandomAccount#49059
United States2140 Posts
| ||
|
onmach
United States1241 Posts
| ||
|
Oxygen
Canada3581 Posts
R1CH: - One man team; - Entirely pro-bono, works in his spare time; - Fixes problems on the fly; - With delay, cast began quicker than MSL; - Personal apology for not predicting everything. R1CH > MSL. Fuck that, R1CH > All. ---- On January 31 2010 10:17 JWD wrote: hahahahahahaha I don't know what's more funny, this post or the fact that jalstar didn't get it QF lol. | ||
|
HazMat
United States17077 Posts
 The casters being out of sync was kind of annoying. Is that possible to fix? | ||
|
CDRdude
United States5625 Posts
| ||
|
crate
United States2474 Posts
| ||
|
dibban
Sweden1279 Posts
| ||
|
000.Zulu
Germany83 Posts
| ||
|
SoL[9]
Portugal1370 Posts
When Artosis and Tasteless said that was 5a.m i tho the delay was because they are still sleeping LOL :D | ||
|
disco
Netherlands1667 Posts
| ||
|
torm3ntin
Brazil2534 Posts
| ||
|
MrShankly
United Kingdom371 Posts
| ||
|
Matoo-
Canada1397 Posts
On January 31 2010 12:05 R1CH wrote: When we go live with something like the TSL, we have people trying to load it in all kinds of confusing ways such as their browser, winamp, realplayer, GOM, etc, combined with F5ing causing many reconnects. The server itself can easily handle thousands of connections so it isn't so much a capacity issue as some combination of people connecting, their video player reconnection behaviour, what initial headers are sent, etc causing some random bug. Just out of curiosity, why not stick to just one browser-embedded Flash or Silverlight stream? It's been a while since I checked up on projects like Red5 but it should be pretty solid nowadays, and it would ensure that everyone uses the same client software.Cost/quality issues? | ||
|
oberon
United States1320 Posts
On January 31 2010 06:51 Liquid`Nazgul wrote: Sorry guys! We try to improve with every cast that's all we can promise at this point. This has been very noticeable, even just within TSL2 -- the production quality of the stream has increased each time I watch. I feel bad for R1ch -- it's always painful when you've got your new code ready to step up and deliver and it...doesn't. Nonetheless, I think everyone on TL has the confidence in R1ch to believe it'll be working soon. --oberon | ||
| ||
