evveloper

What Happened???

ronald — Thu, 07 Aug 2008 06:20:02 +0000

Some people may be asking themselves this very question — what happened? Specifically, whathappened to the road map??? Well, it did not really pan out - but read on - because it is not all that bad.

As most can guess, Evver Games development comes and goes in phases. While I had great intentions of trodding down the path outlined in the roadmap, I found it hard to get motivated to do so. A great deal of the motivation problem was that I was not really satisfied with the technology being used to implement the games. The technology (AJAX) felt slow, fairly limiting, and well… it had just lost interest for me. Long story short, I always found other things to do than work on Evver Games.

Now the good news. Evver Games is actually under active development. It has been for over a month, and I am making fairly good progress. The core libraries are being revamped, becoming much easier to use, and better yet - the front end is getting a complete remake. I have began implementing the games using Flex 3.

Flex looks awesome, very powerful, and easy to use. I have quite a bit of a learning curve, as I know nothing about Flex or Flash; but that learning curve has got me excited about spending time developing Evver Games again. I see a lot of great possibilities, and I also see a much easier road ahead for more games, quicker releases, etc…

I am not going to pre-announce anything, or hype a game that is no where near completion. I am going to leave this as it is, but I hope to be posting more soon. Stay tuned!

Evver Games Roadmap

ronald — Fri, 19 Oct 2007 21:12:12 +0000

I am sure some are interested in getting a more detailed roadmap than what is currently available in the developers section of evver.com. Here are some of my notes for what I would like to see. Naturally, these notes are very brief, as they were only meant to serve me - but this should give you some sense of where Evver Games is headed:

1.0 - Planned
Write developer documentation
Make UML diagrams

0.9 - Planned
Add Logic Game
Add in-game help

0.8 - Planned (Herald Hibernate)
Add Hibernate abilities
Add Leader Board

0.7.1 - Planned (Geeky Galvin)
Improve chat, make option based
Make media project, move images to this
Add background choices (brown leather, green velvet)
Add Memory Game

0.7 - In Development (Freddie Freedom)
Load external resources
Change license to GPL
Release no peeking
Clean up API

0.6.1 - Released July 4, 2007 (Eddie Elegance)
Refine graphics
Center graphics
Switch background each round

0.6 - Released May 31, 2007 (Dewy Decoupler)
Decouple cardservlet from Golf
Refine card platform
Add signal capability
Add Ant build files

0.5.1 - Released January 6, 2007 (Maintenance Release)
Added Chat Feature

0.5 - Released December 26, 2006 (Initial Release)

Ajax Solitaire

ronald — Mon, 27 Aug 2007 03:07:17 +0000

Runs on any browser, any platform. Multiple versions of solitaire available…Very cool!

I can take a lot of hints from this guy. I love what he has done with the site. It is one of the nicest attempts at ajax card games I have evver seen.

ProxyPass with Plesk

ronald — Fri, 24 Aug 2007 06:51:14 +0000

I attempted to implement ProxyPass on a Plesk subdomain. No big deal right? Wrong. At least for me and a slew of other people in Google land. I eventually came up with a clean and simple solution. Its one that I did not see anywhere else, so I thought it worth sharing here. We will get to that soon, but first a bit of background:

After creating my subdomain in Plesk, whenever I tried to load a web page from the subdomain, I would either be forwarded to the main site (my default action for dealing with unknown subdomains ), or I would encounter a 403 Forbidden error. The difference seemed to only be if I included a path in the URL or not. If a path was present, I would get a 403 Forbidden page; if a path was not used, I would be forwarded. In either case, the following message was in the Apache error_log:

[Wed Aug 22 17:49:20 2007] [error] [client <--IP Address--> ] client denied by server
configuration: proxy:http://0.0.0.0:8080/

Many of the solutions that I had been reading online emphasized the need to eliminate any directives not directly related to ProxyPass. Given the state of Plesk, and its desire to completely control the directives found inhttpd.include, there was little hope of doing this. I had tried to get by just using vhost.conf in my subdomain, but this resulted in the same error.

I then decided to delete the subdomain in Plesk, and make my own virtual host in Apache. I created a file (its name is not important), and placed it in the sites-available folder, the linked to it from my sites-enabled folder. This is the standard Debian was to configure Apache, but your installation may vary. The contents of VirtaulHost section is as follows:

ServerName my-subdomain-name.evver.com
ProxyPass / http://localhost:8080/
ProxyPassReverse / http://localhost:8080/

As you can see, the ProxyPass is set to forward requests to another server sitting on port 8080. I have left off ProxyRequests Off, and others, as these are the defaults. You will often see the ProxyRequests Off directive included in other configurations, as it is potentially hazardous to leave open. I knew from other configuration files this was off, so I omitted it from this Virtual host for the sake of brevity.

This will hopefully give you an easy way to implement ProxyPass while not breaking Plesk. If you continue to have issues, you will want to ensure mod_proxy is properly configured for Apache. Remember that proxy.conf, proxy.load, and proxy_http.load all need to be available as modules to Apache. In my Debian environment, this mean simply linking these files from mods-available into mods-enabled. You may also need to edit proxy.conf and allow incoming connections in the … block.

So in the end, one small file led to an easy way to manage ProxyPass with Plesk.

Top 100 Independent Games

ronald — Mon, 20 Aug 2007 23:20:37 +0000

Here is a list of some interesting games. It looks like most are of the ‘downloadable’ variety, but some online games make it into the mix as well. Without looking closely at the games, it appears that some are done by individuals, or small teams. Good work!

From Digg: Independent games are interesting, but it can be hard to find the best games. Without further ado, Game Tunnel presents the Top 100 Indie games from the last 3 years!

Boolean Memory Use

ronald — Tue, 14 Aug 2007 17:37:58 +0000

A friend recently asked: “What’s the deal with [Java] boolean types? Are they pretty compact in memory if I have several in a class, or should I try to use an int and do bitwise operations like I would in C?”

This is a great question, and as I was finishing up an email response, I thought it may be a good entry for the evveloper blog. Hopefully I have provided enough information that you can see what would be best in your situation:

‘boolean’ primitives are treated as an int (when declared standalone) or a byte (when declared in an array) in java. Any primitive that is local to a method is kept on the call stack, and will be cleared when the method falls out of scope. Because it is kept on the call stack, operations are handled quickly. Global and static booleans are kept on the heap, and will be garbage collected (like regular objects), when they are no longer referenced.

‘Boolean’ objects are a bit larger, for reasons like the object header, etc… But every Boolean.TRUE shares the same hashcode and equals(), and so all instances really just reference the same object. This, of course, is also true for Boolean.FALSE. These objects are known as ‘flyweights’ and are considered to have a size of 0 for practical purposes (because java itself will create the first insatnce, and all others after that just reference that instance). Because these objects are kept on the heap, access to them will generally be a bit slower than primitives.

So, you are faced with the classic dilemma of choosing between memory footprint and speed, as is always the case. Storing booleans in an int, and doing bitwise operations is a nice balance between memory and speed, but is discouraged by most, because of issues like compatibility with future versions, software expandability, and code clarity and reuse. The days of super efficient code and have replaced by super reusable (and thus programmer efficient) code, unless there is a strong case not too.

2 Critical VPS Parameters

ronald — Mon, 13 Aug 2007 14:25:22 +0000

DZone readers were recently told to ‘think twice’ about Virtual Private Servers. This reminded me of a previous experience, and the lesson I learned - understanding Virtuozzo parameters. Knowing these parameters can help you avoid what could amount to a 20MHz mistake.

Each Virtuozzo node, or VE, is controlled by a set or parameters and user bean counters. Most of the user bean counters are easy to understand. An UCB such as ‘numproc’ is the number of number of processes that are allowed, ‘numtcpsock’ is the number of TCP sockets allowed, and so forth. These user bean counters and parameters fairly straight forward, except for two very important parameters related to CPU use.

The first parameter, ‘cpulimit’, is actually not hard to understand at all once it has been explained. This parameter denotes the maximum amount of CPU a particular node can use. What can sometimes cause wonder is how you can end up with something like 200%, or even 400%. This is simply the maximum amount of each processor core a node is allowed to use. On a machine which only has one core, the amount can only be 100%, but for a machine with 4 cores, the amount could be as much as 400%.

The other parameter, ‘cpuunits’, was more elusive and harder to find information on. This parameter represents the minimum number of units a node is guaranteed to have. According to the Virtuozzo manual, 1GHz is equivalent to approximately 50,000 ‘cpuunits’. Additionally, the manual states that the number of bogomips for a system, multiplied by 25, will also give the total number of ‘cpuunits’ for a Virtuozzo host. On a Linux system, /proc/cpuinfo will tell you the number of bogomips for each core.

Using the above cpulimit and cpuunits parameters, you can determine the minimum and maximum CPU your VPS node is guaranteed. I did this exercise using both methods on my previous and current VPS. My previous VPS was an AMD 64 X2 4400+, with cpuunits = 1000, and cpulimit = 20%. My current VPS is a Dual core Dual Xeon, with cpuunits = 9,999, and cpulimit = 400%. Here are the results from each method for both my previous and current plans:

Previous VPS, Method 1:
Method: 1GHz = 50,000 cpuunits
Node: 4.4GHz * 50,000 = 220,000 cpuunits
VPS Configuration: 1000 cpuunits / 220,000 total cpuunits = 00.45% of CPU

Previous VPS, Method 2:
Method: Total Bogomips * 25 = total node cpuunits
Node: (4425.70 core 0 + 4422.08 core 1) * 25 = 221,194.5
VPS Configuration: 1000 cpuunits / 221,194.5 total cpuunits = 00.45% of CPU

Current VPS, Method 1:
Method: 1GHz = 50,000 cpuunits
Node: 8.0GHz * 50,000 = 400,000 cpuunits
VPS Configuration: 9999 cpuunits / 400,000 total cpuunits = 02.50% of CPU

Current VPS, Method 2:
Method: Total Bogomips * 25 = total node cpuunits
Node: (3992.49 core 0 + 3989.97 core 1 + 3990.06 core 2 + 3990.03 core 3) * 25 = 399,063.75
VPS Configuration: 9999 cpuunits / 399,063.75 total cpuunits = 02.50% of CPU

The previous VPS only guaranteed one half of 1% CPU!!!! No wonder it was slow! It was only guaranteed to get ~20MHz with an upper limit or burst of ~440MHz. The current VPS is configured to receive at least 2.5% CPU, and up to the full 400%. This places the current VPS CPU somewhere between ~200MHz and 8GHz. The reseller could put as many as 32 VPS accounts on a host if they do not oversell based on RAM. This would mean no one would actually only get ~200MHz on average, only far better, but it is nice to know that I am at least guaranteed that much by the parameters.

Additional Note: I also benchmarked both the previous and current VPS nodes. The results show a huge difference. Take a look.

How fast is Evver?

ronald — Sat, 11 Aug 2007 00:17:11 +0000

I have noticed some people have been using pingdom to test just how fast this virtual server is. I am just fine with that, but I think it should be done properly. Evver.com’s main page can be a bit slow to load at times, but by no fault of it’s own.

I have several javascripts on the main page, some for loading ads, and some just for presentation. While the presentation scripts do not really cause any issue, the ad scripts can be painfully slow depending on how the services I use are doing at that particular time. Thus, sometimes Evver appears slower to load than what is really fair. For this reason, I suggest that if you would like to test, use blog.evver.com as your test url. It is still the same VPS, but with fewer scripts than are on the main page. I think this will give you the most reliable result.

90+ Social Games, AJAX Games and Mashups

ronald — Wed, 08 Aug 2007 22:53:06 +0000

Quote from Mashable.com:

“An established gaming network got bought this week, but we’re far more interested in the newcomers. Lately, many web developers have been taking their games to the next level with ajax, social networking and mashups of…”

This is excatly the kind of thing that Evver.com needs to go after. Being considered part of the new “Online Games 2.0″, whatever that is supposed to mean, must be good… it’s got ‘marketing team’ written all over it!

Seriously though, maybe after looking at some of the games here I can get a good idea about how to use social netowrks, etc… to promote Evver. Any hints / recommendations are welcome!

A Faster Virtual Server

ronald — Mon, 06 Aug 2007 04:49:10 +0000

Here is an article I wrote several months ago for Evver.com. It is no longer posted there, as the content for the site is becoming more concentrated, but it is searched for often enough to repost it here:

Disgruntled with my previous VPS provider, I started looking for a new Virtual Private Server — but this time I used the UnixBench benchmark as my guide. The outcome was a new hosting provider for about the same money, and a 1370% faster Debian VPS. Here is the story:

Mid December, I signed up for VPS Plan 2 with Canaca.com. This plan includes 260MB RAM and 50GB disk space, and would run Debian. For a small web site, it fit my bill well. I would get the feel of dedicated hosting, with the bill of something much cheaper. When the account was activated I was given 614MB RAM and 100GB disk space! Nice, these were the specs of Plan 3! I thought this was going to be a sweet ride. At least I had hoped.

What soon began to happen was ridiculously slow response times! Issuing even a simple ‘ls’ command in a shell would take several seconds! These were small folders too. I emailed tech support, but no meaningful response came. The technician simply said that it behaved as would be expected. Hmmmm…. How was I going to get someone to take a look at this?

WebHostingTalk has a thread comparing VPS web hosting plans from several providers using UnixBench WHT. I decided to give it a shot and see how Canaca stacked up to the others. Here are the results:

TEST	BASELINE	RESULT	INDEX
Dhrystone 2 using register variables	376783.7	1064124.4	28.2
Double-Precision Whetstone	83.1	1429.5	172.0
Execl Throughput	188.3	337.2	17.9
File Copy 1024 bufsize 2000 maxblocks	2672.0	7004.0	26.2
File Copy 256 bufsize 500 maxblocks	1077.0	1844.0	17.1
File Read 4096 bufsize 8000 maxblocks	15382.0	47096.0	30.6
Pipe Throughput	111814.6	150968.7	13.5
Pipe-based Context Switching	15448.6	37634.2	24.4
Process Creation	569.3	851.0	14.9
Shell Scripts (8 concurrent)	44.8	65.9	14.7
System Call Overhead	114433.5	203644.1	17.8
			====
FINAL SCORE			24.0

For anyone familiar with UnixBench WHT, you can see this is not an overwhelmingly great score. In fact, it is pretty pathetic. What followed this test was a number of emails to Canaca, and their lack of a response thereafter. In these emails, I even went into detail on their vps configuration, laying out what I thought were bad cpuunit and cpulimit parameters (but more on those in a future post). This, combined with a number of downtimes, one of which was quite prolonged, compelled me to switch hosting providers.

The same thread on WebHostingTalk led me to slhost.com. Their benchmark numbers were incredible, and other threads had a number of very positive comments for them. The price is normally about 2x as much as Canaca, but I was able to sign up at a locked-in 50% discount. I choose Plan 1, which includes 256MB RAM and 10GB disk space - and requested Debian over the default CentOS. These specs are not quite the same as what I previously had, but it’s really all I need. Here the benchmark results of the new VPS:

TEST	BASELINE	RESULT	INDEX
Dhrystone 2 using register variables	376783.7	25307242.1	671.7
Double-Precision Whetstone	83.1	1390.4	167.3
Execl Throughput	188.3	5912.1	314.0
File Copy 1024 bufsize 2000 maxblocks	2672.0	167299.0	626.1
File Copy 256 bufsize 500 maxblocks	1077.0	47811.0	443.9
File Read 4096 bufsize 8000 maxblocks	15382.0	1164483.0	757.0
Pipe Throughput	111814.6	2729615.7	244.1
Pipe-based Context Switching	15448.6	527627.0	341.5
Process Creation	569.3	14727.0	258.7
Shell Scripts (8 concurrent)	44.8	1606.7	358.6
System Call Overhead	114433.5	2158793.0	188.7
			====
FINAL SCORE			353.7

Nice! Not only do these numbers look good, the server feels good. It is fast, and responds as well as the numbers reflect. For the kind of traffic I receive, it feels like a dedicated host. Now, just to make things clear — I am NOT affiliated with slhost.com in any way, except as a happy customer. I have NOT joined their affiliate program, just so this review will be read without a hint of bias.

So, what does this mean for you? Take a look at your provider using UnixBench WHT. It may prove to be worth your time in a substantial way!

Additional Note: If you would like to test this VPS, please see this post. Thanks.