暂无图片
暂无图片
暂无图片
暂无图片
1
暂无图片
文档 / 2000-07 Eric Brewer: Towards Robust Distributed Systems (PODC'2000 Keynote) (ACID vs. BASE, The CAP Theorem).pdf
2000-07 Eric Brewer: Towards Robust Distributed Systems (PODC'2000 Keynote) (ACID vs. BASE, The CAP Theorem).pdf
小鱼儿
64
12页
0次
2024-01-24
免费下载
1
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Towards Robust
Towards Robust
Distributed Systems
Distributed Systems
Dr. Eric A. Brewer
Dr. Eric A. Brewer
Professor, UC Berkeley
Professor, UC Berkeley
Co
Co
-
-
Founder & Chief Scientist, Inktomi
Founder & Chief Scientist, Inktomi
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Inktomi at a Glance
Inktomi at a Glance
Company Overview
Company Overview
“INKT” on NAS DAQ
“INKT” on NAS DAQ
Founded 1996 out of UC
Founded 1996 out of UC
Berkeley
Berkeley
~700 Employees
~700 Employees
Applications
Applications
Search Technology
Search Technology
Networ k Produc ts
Networ k Produc ts
Online Shopping
Online Shopping
Wireless Systems
Wireless Systems
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Our Perspective
Our Perspective
Inktomi builds two
Inktomi builds two
distributed systems:
distributed systems:
–
–
Global Search Engines
Global Search Engines
–
–
Distributed Web Caches
Distributed Web Caches
Based on scalable
Based on scalable
cluster & parallel
cluster & parallel
computing technology
computing technology
But very little use of
But very little use of
classic DS research...
classic DS research...
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
“
“
Distributed Systems” don’t work...
Distributed Systems” don’t work...
There exist working DS:
There exist working DS:
–
–
Simple protocols: DNS, WWW
Simple protocols: DNS, WWW
–
–
Inktomi search, Content Delivery Networks
Inktomi search, Content Delivery Networks
–
–
Napster, Verisign, AOL
Napster, Verisign, AOL
But these are not classic DS:
But these are not classic DS:
–
–
Not distributed objects
Not distributed objects
–
–
No RPC
No RPC
–
–
No modularity
No modularity
–
–
Complex ones are single owner (except phones)
Complex ones are single owner (except phones)
2
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Three Basic Issues
Three Basic Issues
Where is the state?
Where is the state?
Consistency vs. Availability
Consistency vs. Availability
Understanding Boundaries
Understanding Boundaries
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Where’s the state?
Where’s the state?
(not all locations are equal)
(not all locations are equal)
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Santa Clara Cluster
Santa Clara Cluster
• Very uniform
• No monitors
• No people
• No cables
• Workin g p ower
• Working A/C
• Workin g B W
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Delivering High Availability
Delivering High Availability
We kept up the service through:
We kept up the service through:
Crashes & disk failures (weekly)
Crashes & disk failures (weekly)
Database upgrades (daily)
Database upgrades (daily)
Software upgrades (weekly to monthly)
Software upgrades (weekly to monthly)
OS upgrades (twice)
OS upgrades (twice)
Power outage (several)
Power outage (several)
Network outages (now ha ve 11 connec tions)
Network outages (now ha ve 11 connec tions)
Physical move of all equipment
Physical move of all equipment
(twice)
(twice)
3
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Persistent State is HARD
Persistent State is HARD
Classic DS focus on the computation, not th e data
Classic DS focus on the computation, not th e data
–
–
this is WRONG, computation is the easy part
this is WRONG, computation is the easy part
Data centers exist for a reason
Data centers exist for a reason
–
–
can’t have consistency or availability without them
can’t have consistency or availability without them
Other locations are for caching only:
Other locations are for caching only:
–
–
proxies,
proxies,
basestations
basestations
, set
, set
-
-
top boxes, desktops
top boxes, desktops
–
–
phones,
phones,
PDAs
PDAs
, …
, …
Distributed systems can’t ignore location
Distributed systems can’t ignore location
distinctions
distinctions
AP
Active Proxy:
Bootstraps thin devices
into infrastructure, runs
mobile code
AP
Workstations & PCs
Berkeley Ninja Architecture
Berkeley Ninja Architecture
Base: Scalable, highly-
available platform for
persistent-state services
Internet
PDAs
(e.g. IBM Workpad)
Cellphones, Pagers, etc.
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
Consistency vs. Availability
Consistency vs. Availability
(ACID vs. BASE)
(ACID vs. BASE)
PODC Keynote, July 19, 2000
PODC Keynote, July 19, 2000
ACID vs. BASE
ACID vs. BASE
DBMS research is about ACID (mostly)
DBMS research is about ACID (mostly)
But we forfeit “C” and “I” for availability,
But we forfeit “C” and “I” for availability,
graceful degradation, and performance
graceful degradation, and performance
This tradeoff is fundamental.
This tradeoff is fundamental.
BASE:
BASE:
–
–
B
B
asically
asically
A
A
vailable
vailable
–
–
S
S
oft
oft
-
-
state
state
–
–
E
E
ventual consistency
ventual consistency
of 12
免费下载
basecap
【版权声明】本文为墨天轮用户原创内容，转载时必须标注文档的来源（墨天轮），文档链接，文档作者等基本信息，否则作者和墨天轮有权追究责任。如果您发现墨天轮中有涉嫌抄袭或者侵权的内容，欢迎发送邮件至：contact@modb.pro进行举报，并提供相关证据，一经查实，墨天轮将立刻删除相关内容。

评论

小鱼儿
暂无图片
1年前
评论
暂无图片 0
来源： https://people.eecs.berkeley.edu/~brewer/PODC2000.pdf https://dblp.org/rec/conf/podc/Brewer00.html https://www.semanticscholar.org/paper/cf1c70afbd942ff34595052c3438dc3f50a90167
1年前
暂无图片 点赞
评论