Test Cata Version * Updated Dec 15, 2010 9:20AM CST *

[Rollie]

New version courtesy of akhtar that hopefully removes the delay from 0 returns.

I am not able to test this until tonight, but wanted to put it up for testing.


http://www.warcraftrealms.com/censusmod ... -5.0.2.zip

[FuxieDK]

New version courtesy of akhtar that hopefully removes the delay from 0 returns.

I am not able to test this until tonight, but wanted to put it up for testing.


http://www.warcraftrealms.com/censusmod ... -5.0.2.zip

Manual /whos are listed in the window, but doesn't get saved in the file
But I think 1974ER already reported this in previous test version..

[FuxieDK]

...............and ShiftClicking a name doesn't announce an override of the queue anymore, it just list the player in the window
....and the player isn't saved in the file either..

[akhtar]

...............and ShiftClicking a name doesn't announce an override of the queue anymore, it just list the player in the window
....and the player isn't saved in the file either..

But does it fix the issue with slow scans, and hung queries with 0 results?

[FuxieDK]

...............and ShiftClicking a name doesn't announce an override of the queue anymore, it just list the player in the window
....and the player isn't saved in the file either..

But does it fix the issue with slow scans, and hung queries with 0 results?

I tested at 5:30 this morning, where even 85-level-range was below 50 online people (so no splits), but none of the level-intervals had zero.

[1974ER]

Akhtar, as far I can tell, yes, censuses are running a LOT smoother and no longer get stuck on 0 found results.

Unfortunately, I have had several system crashes (not related to Censusplus), so I have lost a large amount of data.

[Balgair]

Just tried the latest version (I didn't upgrade to the previous one after hearing the slowness complaints) and seems fine to me, fairly normal census times on my main server anyway

[FuxieDK]

Finally, after going through all of Horde and started on Ally, I managed to find a search without any hits:

[18:12:03] Census+: Sending /who r-"Dwarf" c-"Warlock" 85-85
[18:12:09] Census+: Results For: r-"Dwarf" c-"Warlock" 85-85
[18:12:09] Census+: Sending /who r-"Dwarf" c-"Shaman" 85-85
[18:12:14] Census+: Results For: r-"Dwarf" c-"Shaman" 85-85
[18:12:14] Census+: Processing 10 characters.

A single line, saying 0 result, would be nice, but atleast the delay is gone...

[FuxieDK]

New version courtesy of akhtar that hopefully removes the delay from 0 returns.

I am not able to test this until tonight, but wanted to put it up for testing.


http://www.warcraftrealms.com/censusmod ... -5.0.2.zip

I don't know if others have seen the same, but for me, this version have suffered a 50% decrease in speed.

Last version did each /who every 5 sec (3 in 15 sec), however this version does one alternating 7 & 8 sec (2 in 15 sec).

This means that my census takes 50% longer, e.g. 22 min --> 33 min.
These long census-times, prevent me from letting a census run *unattended* on my laptop, as I'm sometimes logged out, for being AFK for 30 minutes, before 85-range (not to mention 1-84) is finished.

[bringoutyourdead]

Numbers:

I'm seeing fairly steady numbers of about 100 character updates per minute
(using the final numbers at end of census run) while playing the 1-20 levels of Worgan and Goblin.

system: Intel Core2 Quad Q6600 @2.4Ghz -5GB ram Vista64

average latency to realms in the 150ms range

[cconover]

Tried latest test version on very small population (Cho'gall Alliance) and larger population, and it seemed to work just fine. I even got a

WARNING: Too many characters matching r-"Blood Elf" c-"Paladin" 85-85 n-i

and it just continued, so seems like it is working fine.

[1974ER]

Fuxie, one thing of note... Blizzard has, according to some sources and my personal experiences, increased the character limits on realms, which partially counts for extended census durations. Even with the improved version 5.0.2, some realms come up with 4400+ characters and take approximately 48 to 52 minutes to finish. Also worth noting, my system is MUCH weaker than Bringoutyourdead's, average latency very similar in the 125 to 180 ms range, roughly.

[FuxieDK]

Latency is primary based on the length of copper (or whatever line you're using) and the quality of the line itself.
You can easily have low latency on a crappy pc or high latency on an ?ber pc..

Anyway, I just wanted to point out the decrease in process-speed, if Rollie or akhtar was able to do anhthing about it.
50% speed-loss is a high price to pay..

Increased realm-limit can't have THAT MUCH to do with it. It was only raised from 4500 to 5000, which isn't that much. Besides, it shouldn't affect the individual WHOs, as they take same time whether it returns 1 or 50 toons. Increasing the limit only makes you run extra census-ranges, they don't "mess" with the individuals..

[1974ER]

Fuxie... you are forgetting one major factor that plays into the equation... the levelling process. Before Cataclysm, the level 80 range was the only one that regularly split a lot... as of right now, when I keep an eye on the verbose mode results... The censuses split on 85, 84, 83, 82, 81, 80 and on heavily populated realms, on 70 and 60 as well.

And not only that do they split, as Cconover reported, they split a LOT. Please remember that going down to "letter level" adds something like 16 more censuses... at only 5 s per /who, just one additional "letter level" split equals over a minute of work. When that starts to happen with BE paladins, NE druids, Tauren druids, almost all human classes... and on 6 different levels instead of one, the minutes start piling up, quickly.

Bringoutyourdead's results resemble mine, we are getting approximately 90 to 100 characters per minute, which is close to what census achieved before the changes Cataclysm brought on.

I suggest that we wait and see how things start working out after more people get to 85 and things start to "settle down" a bit. In the meanwhile Rollie and Akhtar can work on refinement ideas, fixing the manual /who problems and on Rollie's part also addressing the database slowness and the site in general.

One final note... 500 more might not seem like a lot... but if about 500 is split even remotely evenly... it can push 10 more censuses over to the "letter level" split... that, as noted, adds time.

[FuxieDK]

1974ER.... Please read what I write...
I'm *not* talking about how many individual WHO commands that are sent..
I'm talking about the time each WHO takes.

Whether WHO returns 1 or 50 toons, they take the same time. However *that time* have increased 50%.

This have nothing to do with neither latency, nor how fast your pc is..
It's something in CensusPlus that have changed.. Maybe it can't be improved, maybe it can, but Rollie still need to check if something can be done...

[1974ER]

Fuxie, I am sorry, but you are wrong... I just spent time counting during /whos... the result? Excuting a /who took a count of 3 to a count of 9 = The time required per /who is NOT static... and also, it isn't tied to the exact amount of characters found either... /who 50 took a count of 3, 40 characters took a count of 6, 6 were /whoed within a coung of 9, 24 within a count 5. There is no rhyme or reason to the length a /who takes, but it most certainly is NOT the same amount of time every time and the same amount of characters can take a variable amount of time.

Also, couple of small samples, 586 characters, time required 5 minutes 53 seconds...
283 chacters, 2 minutes 24 seconds... 274 characters, 2 minutes 2 seconds... 535 characters, 3 minutes 41 seconds...

So, simply, no... if a decrease of 9 changes time required by 22 seconds and a decrease of 51 by 2 minutes 12 seconds... the rates do not match, not even close. Also... the last of those censuses excuted the largest number of /whos, still taking less time than the first... Why? Different population structures, the first is a very top heavy realm, the last due to "new players" status, a bottom heavy one.

[akhtar]

WRT timing, wholib dynamically modifies the rate that it sends who requests based upon the responsiveness of the server. Blizz purposely throttles /who requests, so wholib tries to predict how long it should wait in between requests, so that the request will succeed on the first try. It has to do this in order to guarantee that each who request is fulfilled.

My experience with this latest version is that it runs quite smoothly and i really don't see a major difference in the amount of time required.

[Kaldran]

Just loaded the beta and did the first test. I had no issues and everything seemed to work great!

["Velen"]
["Alliance"]

Can't speak to any "slower" issues since I normally don't pay attention to it while it is running (verbose is set to off).