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

 
 

System Architecture Driven by Hardware  

2014-11-28 09:42:50|  分类: 默认分类 |  标签: |举报 |字号 订阅

  下载LOFTER 我的照片书  |

The advancement of h/w is no doubt phenomenal, and it has either invalidated existing design principles or caused a redesign to exploit the h/w for speed.  As often said, speed is not an option but a must in business, and so, the tenet "capacity as data, speed as memory and price as disk" is always being reinforced.  For example, with more cores being squeezed into a chip, more memory storage being created, and hardware features such as h/w transactional memory being made commercially available, database people rush to exploit whatever we can to speed up and remove whatever overheads that we could not overcome with the old servers.  However, that makes the design pretty much h/w dependent.  It is ironical to some extent, as all these while, we want the system to be data independent, but we do not mind them to be h/w dependent so long as we gain the speed.   For each new h/w, we may have to re-examine the design principles [1], trying to make the data structures, transaction management etc to be friendly to the new h/w and also to make the system amenable to possibly new optimization.   So, the cycle goes round and round... and this keeps us busy and helps to create space and opportunity for new businesses to emerge...[2]

 

We should take a step back to re-think about the design of DBMS system from s/e view point, so that we do not have to make most operations h/w dependent.  We need to think about the addition of new h/w features.  Is there a protocol or a set of building blocks that facilitates such development?

 

My 2011 blog on "New Architecture for DBMS?"

http://ooibc.blog.163.com/blog/static/103968235201191810110558/

 

References:

[1] K.-L. Tan, Q. Cai, B. C. Ooi, W.F. Wong, C. Yao, H. Zhang: In-memory Databases – Challenges and Opportunities -- From Software and Hardware Perspective. ACM SIGMOD Record, Special Issue on Visionary Ideas in Data Management, June 2015.
[2] H. Zhang, G. Chen, B. C. Ooi, K. L. Tan, M. Zhang: In-Memory Big Data Management and Processing: A Survey. IEEE Transactions on Knowledge and Data Engineering. Vol. 27, No. 7, p1920-1948, 2015

  评论这张
 
阅读(349)| 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