search

4950 / DirtyGame

CSCI 4950 Project
csci4950.azurewebsites.net
3 stars 0 forks source link

Really, 15 second server connections?! #347

Closed Quillwerth closed 10 years ago

Quillwerth commented 10 years ago

Okay, I can't ignore this glaring UX flaw any more. Playing the game is actually PAINFUL because of it. We log a ton of GameEvents, probably because we used to give a damn what tactics people were using. I don't think we care anymore: we care about players, and just raw plays of levels.

Three things I want to know:

  1. What times are the rest of you getting after levels, on average?
  2. What events do we actually still use? I've heard MonsterKilled is important. What about hits?
  3. Do we send all the events on round end? Can we send events live during gameplay instead?
  4. Other ideas for the speed problems? Can we just increase our hardware?
bkern24 commented 10 years ago

Increase hardware!

On Sat, Apr 19, 2014 at 4:22 PM, Patrick Severin notifications@github.comwrote:

Okay, I can't ignore this glaring UX flaw any more. Playing the game is actually PAINFUL because of it. We log a ton of GameEvents, probably because we used to give a damn what tactics people were using. I don't think we care anymore: we care about players, and just raw plays of levels.

Three things I want to know:

  1. What times are the rest of you getting after levels, on average?
  2. What events do we actually still use? I've heard MonsterKilled is important. What about hits?
  3. Do we send all the events on round end? Can we send events live during gameplay instead?
  4. Other ideas for the speed problems? Can we just increase our hardware?

Reply to this email directly or view it on GitHubhttps://github.com/4950/DirtyGame/issues/347 .

Quillwerth commented 10 years ago

Will that bill the university more?

Do I care?

Quillwerth commented 10 years ago

@bkern24 , could you answer the first question? If it's just my internet, then so be it.

Just had a "retrying connection" timeout, twice, despite having internet the whole time. Has anyone else seen that?

bkern24 commented 10 years ago

I hope so.....

On Sat, Apr 19, 2014 at 4:25 PM, Patrick Severin notifications@github.comwrote:

Will that bill the university more?

Do I care?

Reply to this email directly or view it on GitHubhttps://github.com/4950/DirtyGame/issues/347#issuecomment-40881123 .

bkern24 commented 10 years ago

No its really slow too. You can tell because one melee = fast and 10 wall huggers = shoot me

On Sat, Apr 19, 2014 at 4:33 PM, Brandon Kern kernx163@umn.edu wrote:

I hope so.....

On Sat, Apr 19, 2014 at 4:25 PM, Patrick Severin <notifications@github.com

wrote:

Will that bill the university more?

Do I care?

Reply to this email directly or view it on GitHubhttps://github.com/4950/DirtyGame/issues/347#issuecomment-40881123 .

Quillwerth commented 10 years ago

Tonight (hopefully) I'll do a total sweep of the code, looking for every mention of certain event types. Anything that's not being read in by code currently (i. e. is just being logged to the server for "research") will be nixed. @AOEChamp could probably greatly increase my speed here, if he can tell me which events are mission critical.

AOEChamp commented 10 years ago

I don't know off-hand. Check the SQL code. Any events not being used in SQL are probably not critical.

On Sat, Apr 19, 2014 at 6:23 PM, Patrick Severin notifications@github.comwrote:

Tonight (hopefully) I'll do a total sweep of the code, looking for every mention of certain event types. Anything that's not being read in by code currently (i. e. is just being logged to the server for "research") will be nixed. @AOEChamp https://github.com/AOEChamp could probably greatly increase my speed here, if he can tell me which events are mission critical.

Reply to this email directly or view it on GitHubhttps://github.com/4950/DirtyGame/issues/347#issuecomment-40883405 .

AOEChamp commented 10 years ago

The other thing to do is use Fiddler or another sniffer to see exactly what is being sent, how large it is, etc. We can also set it up so that data gets sent asynchronously every few seconds. I initially attempted that but was getting some errors. Might be do-able though.

AOEChamp commented 10 years ago

Furthermore, it seems that azure website policy has an idle timeout period. After the idle period(20 min default, I believe), the site shuts down. The next request will take an extra 10-15s as the site boots up. The policy is not modifiable, so the only solution(if this is actually a problem) is to switch from a website to a Virtual Machine

AOEChamp commented 10 years ago

Hey @fualkner try out the code in my branch. It sends current data async every 5 sec. Should improve end game performance significantly. Let me know if it reduces send time at the end.

Quillwerth commented 10 years ago

This is the best thing ever. I'm not exaggerating when I say 300% faster. I might be under the actual speed increase. Nice, nice, nice work. Once this gets merged, we'll close.