注册 登录  
 加关注
   显示下一条  |  关闭
温馨提示!由于新浪微博认证机制调整,您的新浪微博帐号绑定已过期,请重新绑定!立即重新绑定新浪微博》  |  关闭

Ooi Beng Chin 黄铭钧

Databases, Machine Learning and Systems

 
 
 

日志

 
 

Rightly Solving the Right Problems  

2011-07-21 16:16:57|  分类: 默认分类 |  标签: |举报 |字号 订阅

  下载LOFTER 我的照片书  |

IT has been a change agent in  many business models and has transformed them in many ways. It has also transformed our life to some extent, such as we are spending more time on virtual social networking than networking in real life.  It no doubt also enhances the connection and the quality of life in many ways.  For every new application that is being used in a massive scale, most of database people believe there are database problems to solve. Is this so?  We look at some examples that we may be solving the problems for the wrong reasons.

1) Online games:  As mobile devices become more ubiquitous and wireless techcology develops further, more users will be spending more time on games.  It is therefore important to provide scalable and efficient support for these massive online games. However, most games, especially those shooting games, are action-packed, highly dynamic and fast-paced games, and users play them for the experience and adrenaline rush/excitement.  The users do not care about the accuracy of  each action.  For example, the users will not have the time to check if an enemy is shot 20 times instead of 5 times, as he/she is more worried in anticipating what comes next.   In this case, do we really need transaction management? Do we really care if we have accurately recorded each action?   If not, then database (mini) transaction management has no role to play so long as the state of the game is periodically saved so that it can be recovered should it fail.  How about the scalability? Most online game engines still place a limit on the number of gamers on each instance of the game.  The distributed database people would have wanted to dynamically partition the game space such that each subspace can be managed by a server. However, suppose we do so, the game characters may move from one subspace to another very quickly and frequently, and may query about the situation in nearby subspaces.  This will introduce heavy communication overhead and also frequent merging and splitting of space.  How can then one solve such a problem so that there will not be any restriction on the number of players on a single plane/overlay? This will  improve the game experience, but can we apply what we know from distributed system research onto it?

2) Spatio-temporal Applications: I have seen problems being defined artifically, without considering the real situational and economic dimensions.  For example, suppose we have to build a new fire engine station.  Do we simply build one at a location that minimizes the distance to every other locations that it has to serve?  The issue is much more complicated than this.  Without talking about the economic issues such as the premium of the location, the mix of the tenancy, etc., we have to consider the traffic and road condition,  companies (eg. is it an industrial area?), and many other factors.  Simplifying the problem as a problem of distance minimization problem does not provide any solution to the real world. On the other hand, solving the real problem could be too complex to many, since it involves so many parties (eg. government agencies, town planners,.. ) and issues. 

3) Design centricity: In recent years, the notion that "design" should be considered from day one has been adopted by some disciplines.  In s/w development, can we really consider the design when we design the s/w or algorithm?  There are HCI and cognitive study people who will consider about the user interface and the layout when we manage to cut a useful s/w, but usually, this is done after we have developed something useful.   This is different from trying to provide "schemaless", keyword based, sample based, etc..interfaces to improve the usability of the database.  For such applications, we may need to re-work on the query processing component and logics.

4)...

  评论这张
 
阅读(1007)| 评论(0)
推荐 转载

历史上的今天

在LOFTER的更多文章

评论

<#--最新日志,群博日志--> <#--推荐日志--> <#--引用记录--> <#--博主推荐--> <#--随机阅读--> <#--首页推荐--> <#--历史上的今天--> <#--被推荐日志--> <#--上一篇,下一篇--> <#-- 热度 --> <#-- 网易新闻广告 --> <#--右边模块结构--> <#--评论模块结构--> <#--引用模块结构--> <#--博主发起的投票-->
 
 
 
 
 
 
 
 
 
 
 
 
 
 

页脚

网易公司版权所有 ©1997-2017