104th Dedicated Server Stats

[Cali]

It looks cool,but its to slow..

 

It is likely to be your network connection (transferring the score data) or computer (rendering that data) that is slow. Unfortunately I can't help with these.

 

It's really slow for me also, nice work on the stats.

[yufighter]

And me also, very, very slow. I use opera 11.

[Boberro]

Weird, I reinstalled Java, plugins don't block Java but still I get blank page in Firefox...

 

In IE works, but is terrible slow oO

[Case]

If you get a blank page in Java simply restart your browser and you should see it. Should work with any Java 1.6 version (although I recommend at lest 1.6.0_u10 for hardware accelerated rendering, and 1.6.0_u22 contains a security fix for applets).

I am (and was already) running the latest update 23 (jre-1.6.0_23) on a linux based system and the stats page only shows the empty tabs. I wanted to show a screenshot, but it appears stallturn is down at the moment. So as far as I can tell this isn't a problem with java or the webbrowser, but more likely with the stats program not supporting all architectures. I was able to look at the board using Firefox on Windows.

 

Moa, what is the reason for showing the stats inside Java, and not just upload it to a database and view it through a web browser. MySQL and PHP are very highly optimized for showing large and complex amounts of data, and it seems you are reinventing the wheel by showing it inside Java. I'm not sure, but it may also explain the slow loading of the different pages.

[Moa]

Thanks for your comments Case. It should be fixed, you accessed it while I was doing an update (it's still a beta so gets fix and enhancement updates all the time at the moment). Sorry it was down just as you were accessing it.

 

Since you asked, the reason I use Java is that I personally hate page refreshes for doing simple things. When the number of players gets large (previously around 1000 per month on the 104th) then continually clicking through pages irks me. With Java you can move the columns around (drag n drop), remove some columns (see the icon to the right of the columns on the a2a and a2g pages), sort any column simply by clicking on it and the sorting doesn't require a page refresh, resize windows (especially useful when looking at the various pilot summary pages). You can also use the search and filtering functions to find a pilot that matches a string or show only the pilots that match a string (eg. only display the 51sts pilots on the a2a or a2g pages). All of this also happens on the client's CPU and memory, which means that the system scales to a larger number of simultaneous users all 'dynamically' manipulating their views in near real-time (dynamically changing) - hence the name of the software. "Near real-time" means if you perform an action and click an update button within a few seconds you should see your score change and the details of the event.

 

I wasn't interested in PHP as it requires continual page refreshes and what you can do with a page (dynamic resizing) is limited to what HTML can do - which is a far bit shorter of what you can do in an applet. I tried very hard to do AJAX using the Google Web Toolkit. It is the best AJAX toolkit by far (IMHO) but is still vastly short of the capabilities of an applet. Unfortunately, applets can be problematic (especially as I work through the teething issues - but should settle down very shortly).

 

Thanks once again for your comments. They especially mean something from you given the huge amount of work you did and do with the nice 51st stats pages - you know exactly what it is like getting this stuff going.

 

The slow loading of the pages is very unlikely to be due to Java. Java is faster than C++ and C these days, and pretty close to FORTRAN for raw speed. All the graphics are fully hardware accelerated (as much as possible on the client's machine, unless they have a poor graphics card that is doing software rendering, eg. some Intel chipsets). Most likely it is network bottlenecks (possibly at my end given the number of people accessing it) and the large amount of information being transferred (there are an awful lot of per-pilot statistics that get sent when a pilot selects their page).

 

Incidentally, as I've mentioned before the dynamicscore software makes its statistics fully available though a SOAP webservice. You could use this to get the stats for the 51sts pilots (or anyone else) and merge them with the 51st database. I may also use it to make HTML pages for those platforms that can't use applets (damn Steve Jobs and his decision with the iPhone).

 

The WSDL for the VNAO stats are at:

http://stallturn.com/dynamicscoreweb/DynamicScoreWebService?wsdl

 

The WSDL for the 104th stats is at:

http://stallturn.com/phoenix/DynamicScoreWebService?wsdl

 

You an use SOAPUI (http://www.soapui.org/) and point it at the WSDL to see the supported operations.

 

In the future I may make a REST webservice interface to complement the SOAP one, as that may be easier for some people to deal with (and bookmarking etc will work with it).

 

Edit: While James Gosling is hardly an unbiased commentator regarding Java performance he has a link to a report from INRIA (French research institute) where they analyze Java for supercomputing work and conclude that it is faster than C and nearly as fast as FORTRAN - although networking could be improved. This was back in 2008 and Java has had a lot of optimization since then (eg. all of Java2D is now hardware accelerated using DirectX or OpenGL shaders, if available). Please refer to: http://blogs.sun.com/jag/entry/current_state_of_java_for

Edited December 26, 2010 by Moa
Add link to INRIA report on Java performance

[Case]

Thanks for your comments Case. It should be fixed, you accessed it while I was doing an update (it's still a beta so gets fix and enhancement updates all the time at the moment). Sorry it was down just as you were accessing it.

I just tried it and I still get the empty tabs (see attachment).

 

Since you asked, the reason I use Java is that I personally hate page refreshes for doing simple things. When the number of players gets large (previously around 1000 per month on the 104th) then continually clicking through pages irks me.

But what is the difference between clicking in a Java screen or clicking on a web browser? I agree that with large numbers of players the MySQL queries may take long, but with the right indexing and intelligent design of the tables I keep the load times small.

 

All of this also happens on the client's CPU and memory, which means that the system scales to a larger number of simultaneous users all 'dynamically' manipulating their views in near real-time (dynamically changing)

Does this mean you transfer all the data to the client before you generate the tables? This may explain why the loading is small... With MySQL/PHP the speed depends primarily on the speed of the server hosting the database, and the internet connection between client and server. Since the output from the queries is usually small, the loadtimes are generally determined by the size of the query and the speed of the database server. Since our community is rather small I don't think the load on the database server will ever become problematic.

 

Most likely it is network bottlenecks (possibly at my end given the number of people accessing it) and the large amount of information being transferred (there are an awful lot of per-pilot statistics that get sent when a pilot selects their page).

Again, this is why I opted for the MySQL/PHP approach, as the results of the query are much much smaller than the data in the database.

 

In general the core of my stats program reads the debrief logs and generates all the necessary information, and uploads the data to the MySQL database. All the PHP scripts have to do is to select, sum and count... All the hard work is done in the stats program that is executed every 5 minutes. I do not see the need to do it more often, because when pilots are flying they are not looking at the stats page.

 

While James Gosling is hardly an unbiased commentator regarding Java performance he has a link to a report from INRIA (French research institute) where they analyze Java for supercomputing work and conclude that it is faster than C and nearly as fast as FORTRAN - although networking could be improved.

Working with supercomputers myself, I don't expect to be using Java anytime soon. All our code is generally in C/C++, and since most supercomputers use linux, there is no need for a platform independent language like Java. But as I am sure you recall from your PhD, we scientists generally work differently than computer scientists. For a recent interesting insight: http://arxiv.org/abs/1012.4119.

[Moa]

Interesting points Case. Thanks for taking the time to make them.

[Case]

Interesting points Case. Thanks for taking the time to make them.

I hope you don't see them as criticism. I am merely trying to get your views on this matter, and providing my own. They are just two roads leading to Rome.

[Moa]

Even if they were criticism, they'd still be welcome.

[RIPTIDE]

Moa is a Java Zealot. Leave him alone Case. ;)

[Sharkster64]

I think the stat page is outstanding even if it is using java. It isn't slow on my end. I think the other servers need to consider using this format in their servers. This has got to be the most detailed stat system I have seen yet. I particularily like the "who is online" feature. You the man Moa! :thumbup:

[Moa]

I think the stat page is outstanding even if it is using java. It isn't slow on my end. I think the other servers need to consider using this format in their servers. This has got to be the most detailed stat system I have seen yet. I particularily like the "who is online" feature. You the man Moa! :thumbup:

 

Thanks Sharkster64.

[Distiler]

I like it a lot! but it's super slow to load:

 

-1:45 since clicking the web link to show the air-to-air default page

 

-35 sec to open Riptide's statistics

 

Also, team statistics page doesen't show anything

 

Using a 3Ghz 2-core system with 4GB of ram in a W7 64bit with Google Chrome last version and a 15/1.5 DSL.

Edited December 27, 2010 by Distiler

[Sharkster64]

19 seconds to load the stat page for me. 16 seconds to load a player. Maybe your internet is slow.

[Distiler]

19 seconds to load the stat page for me. 16 seconds to load a player. Maybe your internet is slow.

 

Nope, not slow at all. Big bandwith and low latency. No packet loose at that site. Computer is optimized at all times.

 

PD: right now I've tried again and it was faster, only took like 30 secs. But I tried during all week and it was like more than a minute, right now it's the faster I've seen. Anyway, should be up in half that time. According to resource monitor the server is sending the information to me at only 30KB/s, perhaps information is fragmented or sent sequentally in tiny packets, I don't know I'm not a technician :(

Edited December 27, 2010 by Distiler

[MoGas]

This statsboard is just great, the best stats what I have seen and worked with. Thanks again Moa, keep it up, you cant make everyone happy, keep in mind!

 

:)

[Moa]

Nope, not slow at all. Big bandwith and low latency. No packet loose at that site. Computer is optimized at all times.

 

PD: right now I've tried again and it was faster, only took like 30 secs. But I tried during all week and it was like more than a minute, right now it's the faster I've seen. Anyway, should be up in half that time. According to resource monitor the server is sending the information to me at only 30KB/s, perhaps information is fragmented or sent sequentally in tiny packets, I don't know I'm not a technician :(

 

Thanks for quantifying the timing information Distiller. The information is coming from my server in New Zealand to you in Spain. It's probably not a fast route there (shared with a lot of other traffic from NZ to Europe and vice versa) and you're probably not the only gettting data from the server (which only has 2 MB/s up). traceroute would be interesting. Of course, latency is not the same as bandwidh.

 

Once the stats have been sorted out (they're still in a bug-fix beta stage) they could get hosted somewhere with a higher upload speed. That may help a little.

 

Like I said to another poster though, I can't do too much about the World's network speeds in software. There are just a huge amount of statistics to get across the wire.

[Boberro]

I had troubles viewing stats in Firefox, so I created standalone shortcut and I run stats directly in Java as EXE program :)

Stats are cool, are detailed so much.

[Moa]

Very clever Bob. Just be careful that you get the latest version from time to time. Soon I'll try and load the Java 'WebStart' JNLP file that allows you to run the program locally, but checks the network to see if there is a new version you could download (best of both worlds).

[Pilotasso]

Were the stats frozen for debug? They show old data only now.

[104th_Crunch]

There is a small issue with stats where they need to be manually updated every so often. Moa is on it and it should be fixed soon.

[Moa]

Some pilots have requested that the 104th Dedicated Server pilot statistics be reset as they were experimenting with how the server works and their battle efficiency does ot reflect regular operations.

 

We are allowing our server guests to choose whether to reset the statistics on 2011-01-01 00:00:00 UTC or to keep the 2010 statistics as part of each pilot's career statistics.

 

You can cast your vote to keep the 104th statistics for 2010 or to start with a clean slate in the New Year. As Burger King says, "Have it your way":

 

http://www.104thphoenix.com/modules.php?name=Forums&file=viewtopic&t=2815

[Grimes]

Why not filter it over a set period of time? I see no logical reason to delete stats from 2010 just for the sake of having a clean slate. Furthermore deleting flight records based on players intent is simply a bad idea. Oh you were drunk or screwing around so your stats were impacted negatively? Well who's fault is that? The whole point of global statistics is that it logs everything you've done in the game regardless of your intent, mood, or other factors that have influence on how well you play. Besides, as play-time adds up the effect of having an "off night" tends to average out and become a non-issue.

[Boberro]

I vote for non-delete ;]

[Sharkster64]

I say that the stats get reset after all the bugs are ironed out. I think that it is kind of pointless to reset the stats when we still have bugs in the stat system. Then we will just have to reset again when they do get fixed. That will make people more upset.

 

I am all for resetting after that. :thumbup: