当前在线人数18678
IBM开发出预测服务器崩溃的软件 - 未名空间精华区
首页 - 版面精华区 - 新闻中心精华区 - 科技新闻版精华区 - 精华区文章阅读 首页
未名交友
[更多]
[更多]
IBM开发出预测服务器崩溃的软件

发信人: devil (闲人), 信区: ITnews
标  题: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Fri Jan 26 11:12:59 2001), 站内信件

http://www.sina.com.cn 2001/01/22 13:29 天极网

  世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
器临近崩溃的时候自动安全关闭服务器。

  IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
全关机,防患于蔚然。

  目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
术称做“软件再生”技术。(竹子编译)



--
        If something you can not make it.... fake it!

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 204.69.198.34]
发信人: devil (闲人), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Fri Jan 26 11:15:24 2001), 站内信件

【 在 devil (闲人) 的大作中提到: 】
: http://www.sina.com.cn 2001/01/22 13:29 天极网
:   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: 器临近崩溃的时候自动安全关闭服务器。
:   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: 全关机,防患于蔚然。
donno whether "shut down a server" was consider a "患"?

:   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: 术称做“软件再生”技术。(竹子编译)


--
        If something you can not make it.... fake it!

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 204.69.198.34]
发信人: pjones (彼得), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Fri Jan 26 14:23:00 2001), 站内信件

【 在 devil (闲人) 的大作中提到: 】
: http://www.sina.com.cn 2001/01/22 13:29 天极网
:   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: 器临近崩溃的时候自动安全关闭服务器。
:   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: 全关机,防患于蔚然。
:   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: 术称做“软件再生”技术。(竹子编译)

IBM is sick.  What if this software is over-smart and shutdown the machine
even though the sysadmin doesn't want a shutdown.   AI, AI.  AI means
machine will act by itself sometime, and very likely sometime it is against
user's will.  Applying AI tech in fault-tolerance technology is kind of crazy.
--
AND IF I TOLD YOU THAT I LOVED YOU,
YOU'D MAYBE THINK THERE'S SOMETHING WRONG.
I'M NOT A MAN OF TOO MANY FACES.
THE MASK I WEAR IS ONE.

※ 修改:.pjones 于 Jan 26 14:23:26 修改本文.[FROM: 131.179.64.111]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 131.179.64.111]
发信人: weijing (味精号巡洋舰), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 00:26:47 2001), 站内信件


【 在 pjones (彼得) 的大作中提到: 】
: 【 在 devil (闲人) 的大作中提到: 】
: : http://www.sina.com.cn 2001/01/22 13:29 天极网
: :   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: : 器临近崩溃的时候自动安全关闭服务器。
: :   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: : 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: : 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: : 全关机,防患于蔚然。
: :   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: : 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: : 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: : 术称做“软件再生”技术。(竹子编译)
: IBM is sick.  What if this software is over-smart and shutdown the machine
: even though the sysadmin doesn't want a shutdown.   AI, AI.  AI means
: machine will act by itself sometime, and very likely sometime it is against
: user's will.  Applying AI tech in fault-tolerance technology is kind of crazy.

don't want to argue with you, just want to point out that this sick
company you mentioned happens to have 5 Nobel Prizes, 7 Medals of
Technology, 30000 patents world wide, and most likely you are using
storage that originated from IBM's magnetic disk or rewritable optical
storage inventions; memory using IBM's singale transistor dynamic memory
cell invention...

anyway, i have no problem for using automation to reboot "troubled" servers,
in fact, i believe this could be a useful feature in IBM's popular Tivoli
management software suite..dunno why you are so afraid of smarter computers

--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.97.182.45]
发信人: dfehHo (ehHo), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 01:17:58 2001) WWW-POST


: : IBM is sick.  What if this software is over-smart and shutdown the machine
: : even though the sysadmin doesn't want a shutdown.   AI, AI.  AI means
: : machine will act by itself sometime, and very likely sometime it is against
: : user's will.  Applying AI tech in fault-tolerance technology is kind of crazy.
:
: don't want to argue with you, just want to point out that this sick
: company you mentioned happens to have 5 Nobel Prizes, 7 Medals of
: Technology, 30000 patents world wide, and most likely you are using
: storage that originated from IBM's magnetic disk or rewritable optical
: storage inventions; memory using IBM's singale transistor dynamic memory
: cell invention...

... so ?

Think about Lucent, Bell Lab has more patents than IBM ...
but it just
announced layoff 16K employees.

IBM might be okay, I don't know,  but AIX really sucks, no
wonder IBM
needs Linux for its server products.

:
: anyway, i have no problem for using automation to reboot "troubled" servers,
: in fact, i believe this could be a useful feature in IBM's popular Tivoli
: management software suite..dunno why you are so afraid of smarter computers
:



--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: ]
发信人: SandyG (抽刀断水), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 02:01:18 2001) WWW-POST

That's just a minor improvement on server technology. No
matter whether a
server is abruptly broken down or safely shut down, the key
is that it
is DOWN. It violates the most important role of server
design: "Server
Nerver Dies"! So if they can do a server which is six-nine
reliable, this
technology will never be used.
AI? Hope MS don't use this in their products. Otherwise my
computer will
auto-shutdown every five minutes. "She" is just too smart to
work overtime.
:-)

【 在 devil (闲人) 的大作中提到: 】
: http://www.sina.com.cn 2001/01/22 13:29 天极网
:
:   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: 器临近崩溃的时候自动安全关闭服务器。
:
:   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: 全关机,防患于蔚然。
:
:   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: 术称做“软件再生”技术。(竹子编译)
:
:
:


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 47.102.144.53]
发信人: weijing (味精号巡洋舰), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 02:16:45 2001), 站内信件


【 在 dfehHo (ehHo) 的大作中提到: 】
: : don't want to argue with you, just want to point out that this sick
: : company you mentioned happens to have 5 Nobel Prizes, 7 Medals of
: : Technology, 30000 patents world wide, and most likely you are using
: : storage that originated from IBM's magnetic disk or rewritable optical
: : storage inventions; memory using IBM's singale transistor dynamic memory
: : cell invention...
: ... so ?
: Think about Lucent, Bell Lab has more patents than IBM ...

no, IBM has the most patents, in fact, IBM spends much more in R&D
and produces much more patents than Bell Lab...and in much more diverse
fields like storage, semiconductor, voice recognition, wireless, etc...

: but it just
: announced layoff 16K employees.

IBM is still agressively hiring, in fact, in NYC, we host interview
sessions every other week...compared to Lucent, IBM has much better
management, and its business is much more diverse, thus not much affected
by economy slow down, you can reference to IBM and LU's stock price

: IBM might be okay, I don't know,  but AIX really sucks, no
: wonder IBM
: needs Linux for its server products.

AIX is never aimed at regular consumers, enterprises like it becuz of
its stability...yet this "sucky" OS gengerated billions of dollars of
profit for IBM...actually OS390 is probably a bigger cashcow for IBM,
it is probably "suckier" to you, yet it runs on mission critical
mainframes all over the world.

IBM recognized the potential of Linux, that is why we endorse it and help
it get into mainstream of enterprise computing...Linux is perfect for
mid-range e-servers, there are a lot of benchmarks out there that can
prove it.  IBM embraces Linux earlier, getting a competitive advantage
early, it is mutual beneficial.

: : anyway, i have no problem for using automation to reboot "troubled" servers,
: : in fact, i believe this could be a useful feature in IBM's popular Tivoli
: : management software suite..dunno why you are so afraid of smarter computers


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.97.182.45]
发信人: weijing (味精号巡洋舰), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 02:28:07 2001), 站内信件

the idea is we humans are lazy, say, I own a server placed in a
data center, I definitely do not want to travel a long way to
server farm just to flip a switch, instead, it is better that the
server reboots itself everytime it crashes.

【 在 SandyG (抽刀断水) 的大作中提到: 】
: That's just a minor improvement on server technology. No
: matter whether a
: server is abruptly broken down or safely shut down, the key
: is that it
: is DOWN. It violates the most important role of server
: design: "Server
: Nerver Dies"! So if they can do a server which is six-nine
: reliable, this
: technology will never be used.
: AI? Hope MS don't use this in their products. Otherwise my
: computer will
: auto-shutdown every five minutes. "She" is just too smart to
: work overtime.
: :-)
: 【 在 devil (闲人) 的大作中提到: 】
: : http://www.sina.com.cn 2001/01/22 13:29 天极网
: :   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: : 器临近崩溃的时候自动安全关闭服务器。
: :   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: : 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: : 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: : 全关机,防患于蔚然。
: :   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: : 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: : 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: : 术称做“软件再生”技术。(竹子编译)


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.97.182.45]
发信人: dapengz (Hello World), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 02:32:00 2001), 站内信件

this thing will probably be used in a cluster/replicated environment,
where another machine will take over when the software decides that
a machine is going to fail.  it is very useful since it will increse
the performance of the cluster since it is very hard to take over
a unscheduled failure.

also, even a six nine system will crash.  it just does that very rarely.
and where did you get that "server never dies" philosphy?  the user
does not care whether a server dies or not.  all he/she cares is
whether there is service interruption, data corruption or other
bad things.

microsoft uses lots of AI techniques in the operating system (for
example, the printer trouble shoot guide uses a simple baysian
network.) 

【 在 SandyG (抽刀断水) 的大作中提到: 】
: That's just a minor improvement on server technology. No
: matter whether a
: server is abruptly broken down or safely shut down, the key
: is that it
: is DOWN. It violates the most important role of server
: design: "Server
: Nerver Dies"! So if they can do a server which is six-nine
: reliable, this
: technology will never be used.
: AI? Hope MS don't use this in their products. Otherwise my
: computer will
: auto-shutdown every five minutes. "She" is just too smart to
: work overtime.
: :-)
: 【 在 devil (闲人) 的大作中提到: 】
: : http://www.sina.com.cn 2001/01/22 13:29 天极网
: :   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: : 器临近崩溃的时候自动安全关闭服务器。
: :   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: : 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: : 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: : 全关机,防患于蔚然。
: :   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: : 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: : 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: : 术称做“软件再生”技术。(竹子编译)


--
There are three kinds of persons on this planet.  Those who can count and
those who can't.

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 128.12.194.52]
发信人: microbe (奇妙的微生物), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 02:34:42 2001), 站内信件


Just like Linux's OOM killer, I don't think this Crash Detector
can be relied on.  Will this software be running on the same server?
What if itself dies?  You can't expect to have a reliable software
on an unreliable machine.

Also, rebooting a server sometimes requires human intervention,
such as disk problems.

【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: the idea is we humans are lazy, say, I own a server placed in a
: data center, I definitely do not want to travel a long way to
: server farm just to flip a switch, instead, it is better that the
: server reboots itself everytime it crashes.
: 【 在 SandyG (抽刀断水) 的大作中提到: 】
: : That's just a minor improvement on server technology. No
: : matter whether a
: : server is abruptly broken down or safely shut down, the key
: : is that it
: : is DOWN. It violates the most important role of server
: : design: "Server
: : Nerver Dies"! So if they can do a server which is six-nine
: : reliable, this
: : technology will never be used.
: : AI? Hope MS don't use this in their products. Otherwise my
: : computer will
: : auto-shutdown every five minutes. "She" is just too smart to
: : work overtime.


--
我在门外坐了两天两夜,看着天空在不断的变化。
我才发现,虽然我到这里很久,却从来没有看清楚这片沙漠。

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 160.39.200.28]
发信人: pjones (彼得), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 04:34:08 2001), 站内信件

【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: 【 在 pjones (彼得) 的大作中提到: 】
: : IBM is sick.  What if this software is over-smart and shutdown the machine
: : even though the sysadmin doesn't want a shutdown.   AI, AI.  AI means
: : machine will act by itself sometime, and very likely sometime it is against
: : user's will.  Applying AI tech in fault-tolerance technology is kind of crazy.
: don't want to argue with you, just want to point out that this sick
: company you mentioned happens to have 5 Nobel Prizes, 7 Medals of
: Technology, 30000 patents world wide, and most likely you are using
: storage that originated from IBM's magnetic disk or rewritable optical
: storage inventions; memory using IBM's singale transistor dynamic memory
: cell invention...
Concerning Nobel prize, I guess Lucent has more Nobel prizes than IBM does.
Does that prove anything? Not really.  Lucent is currently a big loser.


It doesn't make sense to honor a ping-pong champion a swimming title.
IBM is good at hardware development because they have tons of competitive
guys working on numerous projects along that direction.  However, their
software sucks---in every software field they lost to somebody: OS2 losts
to Windoz, AIX losts to Solaris and Linux, DB2 losts to Oracle.  What a
loser.  I happened to know some of their AI guys, don't see any brilliance.

: anyway, i have no problem for using automation to reboot "troubled" servers,
: in fact, i believe this could be a useful feature in IBM's popular Tivoli
: management software suite..dunno why you are so afraid of smarter computers

If your smart colleague did something last night and you found your computer
acted weird this morning, guess you wouldn't feel good even if the colleague
claims he is trying to help.  If the computer happens to be a file server
and your boss just scolded you because he never ordered you to shutdown
the computer at 12:34am, then it is a more interesting scenario.
--
AND IF I TOLD YOU THAT I LOVED YOU,
YOU'D MAYBE THINK THERE'S SOMETHING WRONG.
I'M NOT A MAN OF TOO MANY FACES.
THE MASK I WEAR IS ONE.

※ 修改:.pjones 于 Jan 27 04:35:07 修改本文.[FROM: 164.67.35.192]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 164.67.35.192]
发信人: dfehHo (ehHo), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 12:32:13 2001) WWW-POST

【 在 devil (闲人) 的大作中提到: 】
: http://www.sina.com.cn 2001/01/22 13:29 天极网
:
:   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: 器临近崩溃的时候自动安全关闭服务器。

Tandem nonstop server implemented it 20 years ago, that's
why we never
heard of Nasdaq got hacked/crashed.

:
:   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: 全关机,防患于蔚然。
:
:   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: 术称做“软件再生”技术。(竹子编译)
:
:
:


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: ]
发信人: dfehHo (ehHo), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 12:42:40 2001) WWW-POST


:
: AIX is never aimed at regular consumers, enterprises like it becuz of
: its stability...yet this "sucky" OS gengerated billions of dollars of
: profit for IBM...actually OS390 is probably a bigger cashcow for IBM,

who is the biggest customer of those mainframes ? IBM ! what
a joke !!


: it is probably "suckier" to you, yet it runs on mission critical
: mainframes all over the world.
:
: IBM recognized the potential of Linux, that is why we endorse it and help
: it get into mainstream of enterprise computing...Linux is perfect for
: mid-range e-servers, there are a lot of benchmarks out there that can
: prove it.  IBM embraces Linux earlier, getting a competitive advantage
: early, it is mutual beneficial.
:
:


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: ]
发信人: dfehHo (ehHo), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 12:57:40 2001) WWW-POST


: : Think about Lucent, Bell Lab has more patents than IBM ...
:
: no, IBM has the most patents, in fact, IBM spends much more in R&D
: and produces much more patents than Bell Lab...and in much more diverse
: fields like storage, semiconductor, voice recognition, wireless, etc...

Take a look at http://www.bell-labs.com/history/

it shows:

  Bell Labs Facts:
  ...
  Its employees have generated more than 40,000 inventions
since 1925.
  ...
http://www.bell-labs.com/history/75/facts.html


and Lucent has:

  6 Nobel Prizes in Physics shared by 11 scientists
  9 U.S. Medals of Science
  6 U.S. Medals of Technology
  1 Draper Prize
  7 C&C Prizes shared by 12 scientists and engineers
http://www.bell-labs.com/history/heritage/awards.html



--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: ]
发信人: weijing (味精号巡洋舰), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 14:33:43 2001), 站内信件


【 在 pjones (彼得) 的大作中提到: 】
: 【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: : don't want to argue with you, just want to point out that this sick
: : company you mentioned happens to have 5 Nobel Prizes, 7 Medals of
: : Technology, 30000 patents world wide, and most likely you are using
: : storage that originated from IBM's magnetic disk or rewritable optical
: : storage inventions; memory using IBM's singale transistor dynamic memory
: : cell invention...
: Concerning Nobel prize, I guess Lucent has more Nobel prizes than IBM does.
: Does that prove anything? Not really.  Lucent is currently a big loser.

You are right, past doesn't predict future

: It doesn't make sense to honor a ping-pong champion a swimming title.
: IBM is good at hardware development because they have tons of competitive
: guys working on numerous projects along that direction.  However, their
: software sucks---in every software field they lost to somebody: OS2 losts
: to Windoz, AIX losts to Solaris and Linux, DB2 losts to Oracle.  What a
: loser.  I happened to know some of their AI guys, don't see any brilliance.

Concensus shows OS2 was technically superior than win95, DB2 beats Oracle
as well at least on #transactions/s benchmark, IBM wasn't known for
its marketing, but things will change as IBM starts agressive advertiseing.

: : anyway, i have no problem for using automation to reboot "troubled" servers,
: : in fact, i believe this could be a useful feature in IBM's popular Tivoli
: : management software suite..dunno why you are so afraid of smarter computers
: If your smart colleague did something last night and you found your computer
: acted weird this morning, guess you wouldn't feel good even if the colleague
: claims he is trying to help.  If the computer happens to be a file server
: and your boss just scolded you because he never ordered you to shutdown
: the computer at 12:34am, then it is a more interesting scenario.

The point is we are talking about smart AI that correctly detect system
failure, most likely on server clusters in enterprise environment

--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.97.182.42]
发信人: weijing (味精号巡洋舰), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 14:37:30 2001), 站内信件


【 在 dfehHo (ehHo) 的大作中提到: 】
: 【 在 devil (闲人) 的大作中提到: 】
: : http://www.sina.com.cn 2001/01/22 13:29 天极网
: :   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: : 器临近崩溃的时候自动安全关闭服务器。
: Tandem nonstop server implemented it 20 years ago, that's

if the server was to be non-stop, then why it has auto-reboot feature?

: why we never
: heard of Nasdaq got hacked/crashed.

wrong, Nasdaq crashed many time, I already heard a couple of times
last year alone.  Maybe you don't watch stock market long enough.

: :   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: : 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: : 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: : 全关机,防患于蔚然。
: :   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: : 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: : 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: : 术称做“软件再生”技术。(竹子编译)


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.97.182.42]
发信人: weijing (味精号巡洋舰), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 14:39:35 2001), 站内信件


【 在 dfehHo (ehHo) 的大作中提到: 】
: : AIX is never aimed at regular consumers, enterprises like it becuz of
: : its stability...yet this "sucky" OS gengerated billions of dollars of
: : profit for IBM...actually OS390 is probably a bigger cashcow for IBM,
: who is the biggest customer of those mainframes ? IBM ! what
: a joke !!

Wrong again, mainframe's biggest customers are corporations and
governments all over the world...do you know China is a big customer
of AS390?

: : it is probably "suckier" to you, yet it runs on mission critical
: : mainframes all over the world.
: : IBM recognized the potential of Linux, that is why we endorse it and help
: : it get into mainstream of enterprise computing...Linux is perfect for
: : mid-range e-servers, there are a lot of benchmarks out there that can
: : prove it.  IBM embraces Linux earlier, getting a competitive advantage
: : early, it is mutual beneficial.


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.97.182.42]
发信人: weijing (味精号巡洋舰), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 14:48:33 2001), 站内信件

Well, well, I guess "patents" mean inventions that make money.

Since Lucent is in such bad shape, I don't see why not more and more
Bell Lab reseachers go work for IBM Research, for better pay and
better benefits at least.

【 在 dfehHo (ehHo) 的大作中提到: 】
: : no, IBM has the most patents, in fact, IBM spends much more in R&D
: : and produces much more patents than Bell Lab...and in much more diverse
: : fields like storage, semiconductor, voice recognition, wireless, etc...
: Take a look at http://www.bell-labs.com/history/
: it shows:
:   Bell Labs Facts:
:   ...
:   Its employees have generated more than 40,000 inventions
: since 1925.
:   ...
: http://www.bell-labs.com/history/75/facts.html
: and Lucent has:
:   6 Nobel Prizes in Physics shared by 11 scientists
:   9 U.S. Medals of Science
:   6 U.S. Medals of Technology
:   1 Draper Prize
:   7 C&C Prizes shared by 12 scientists and engineers
: http://www.bell-labs.com/history/heritage/awards.html


--
※ 修改:.weijing 于 Jan 27 14:53:26 修改本文.[FROM: 32.97.182.42]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.97.182.42]
发信人: dfehHo (ehHo), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 16:31:44 2001) WWW-POST

【 在 weijing (味精号巡洋舰) 的大作中提到: 】
:
: 【 在 dfehHo (ehHo) 的大作中提到: 】
: : 【 在 devil (闲人) 的大作中提到: 】
: : : http://www.sina.com.cn 2001/01/22 13:29 天极网
: : :   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: : : 器临近崩溃的时候自动安全关闭服务器。
: : Tandem nonstop server implemented it 20 years ago, that's
:
: if the server was to be non-stop, then why it has auto-reboot feature?
:
: : why we never
: : heard of Nasdaq got hacked/crashed.
:
: wrong, Nasdaq crashed many time, I already heard a couple of times
: last year alone.  Maybe you don't watch stock market long enough.

That was internet bubble bursted.

:
: : :   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: : : 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: : : 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: : : 全关机,防患于蔚然。
: : :   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: : : 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: : : 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: : : 术称做“软件再生”技术。(竹子编译)
:
:


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: ]
发信人: devil (闲人), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 16:33:40 2001), 站内信件

【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: 【 在 pjones (彼得) 的大作中提到: 】
: : Concerning Nobel prize, I guess Lucent has more Nobel prizes than IBM does.
: : Does that prove anything? Not really.  Lucent is currently a big loser.
: You are right, past doesn't predict future
: : It doesn't make sense to honor a ping-pong champion a swimming title.
: : IBM is good at hardware development because they have tons of competitive
: : guys working on numerous projects along that direction.  However, their
: : software sucks---in every software field they lost to somebody: OS2 losts
: : to Windoz, AIX losts to Solaris and Linux, DB2 losts to Oracle.  What a
: : loser.  I happened to know some of their AI guys, don't see any brilliance.
: Concensus shows OS2 was technically superior than win95, DB2 beats Oracle
: as well at least on #transactions/s benchmark, IBM wasn't known for
according to Oracle's ad, you will be rich by prove this ..,. :-)
kidding...

anyway, IBM looks in good shape recently..
pretty aggressive and fast-reacting ..

IBM & Bell-Lab are two companies deserve our respect.

: its marketing, but things will change as IBM starts agressive advertiseing.
: : If your smart colleague did something last night and you found your computer
: : acted weird this morning, guess you wouldn't feel good even if the colleague
: : claims he is trying to help.  If the computer happens to be a file server
: : and your boss just scolded you because he never ordered you to shutdown
: : the computer at 12:34am, then it is a more interesting scenario.
: The point is we are talking about smart AI that correctly detect system
: failure, most likely on server clusters in enterprise environment


--
        If something you can not make it.... fake it!

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 204.69.198.34]
发信人: zamster (彼岸花), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 16:34:01 2001) WWW-POST

【 在 dfehHo (ehHo) 的大作中提到: 】
:
: :
: : AIX is never aimed at regular consumers, enterprises like it becuz of
: : its stability...yet this "sucky" OS gengerated billions of dollars of
: : profit for IBM...actually OS390 is probably a bigger cashcow for IBM,
:
: who is the biggest customer of those mainframes ? IBM ! what
: a joke !!
:
Is that a question or a comment? No, IBM is not the biggest
customer of it's own product:) After all, how many
mainframes do you need for one shop? They are $2 millilion
to $5 million a piece.  And IBM shiped a few hundreds of
them (the newset Z900 model alone) in fourth querter.

So what's the joking matter you are seeing?



--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: adsl-138-89-99-]
发信人: zamster (彼岸花), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 16:48:09 2001) WWW-POST

【 在 pjones (彼得) 的大作中提到: 】
: It doesn't make sense to honor a ping-pong champion a swimming title.
: IBM is good at hardware development because they have tons of competitive
: guys working on numerous projects along that direction.  However, their
: software sucks---in every software field they lost to somebody: OS2 losts
: to Windoz, AIX losts to Solaris and Linux, DB2 losts to Oracle.  What a
: loser.  I happened to know some of their AI guys, don't see any brilliance.
:
If this is a technology discussion, then don't bring
marketing factor into it. IBM's marketing has contributed to
a few failures, and they are catching up. Here's nice little
story about Microsoft, do you know Microsoft runs their
business on IBM mainframe? well, AS400, not exactly
mainframe yet.

At one point Microsoft had 23 AS/400s located around the
world. It was a very good customer, went from CISC to RISC,
and was always one of the first to upgrade to new
technology. Then came the Year 2000 problem, and despite
five years of dedicated service during a period of great
revenue growth, the company decided that it was time to
move off the AS/400. So in June of 1999, the company
unplugged its AS/400s and powered up 1200 NT servers it
needed to replace them. But things didn't quite go as
planned. They found they couldn't make it work. So last
year, one year after unplugging their AS/400s, they're
back on the AS/400.

Can you think of a company with greater incentive to move to
NT, and they couldn't do it?




--
※ 修改:·zamster 於 Jan 27 16:48:09 修改本文·[FROM: adsl-138-89-99-]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: adsl-138-89-99-]
发信人: zamster (彼岸花), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 17:30:00 2001) WWW-POST

【 在 devil (闲人) 的大作中提到: 】
: 【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: : 【 在 pjones (彼得) 的大作中提到: 】
: : : Concerning Nobel prize, I guess Lucent has more Nobel prizes than IBM does.
: : : Does that prove anything? Not really.  Lucent is currently a big loser.
: : You are right, past doesn't predict future
: : : It doesn't make sense to honor a ping-pong champion a swimming title.
: : : IBM is good at hardware development because they have tons of competitive
: : : guys working on numerous projects along that direction.  However, their
: : : software sucks---in every software field they lost to somebody: OS2 losts
: : : to Windoz, AIX losts to Solaris and Linux, DB2 losts to Oracle.  What a
: : : loser.  I happened to know some of their AI guys, don't see any brilliance.
: : Concensus shows OS2 was technically superior than win95, DB2 beats Oracle
: : as well at least on #transactions/s benchmark, IBM wasn't known for
: according to Oracle's ad, you will be rich by prove this ..,. :-)
: kidding...
:
: anyway, IBM looks in good shape recently..
: pretty aggressive and fast-reacting ..
:
: IBM & Bell-Lab are two companies deserve our respect.
Correct, Watson lab and Bell lab are two legendary research
centers that deserve a lot of respect. It makes no sense to
argue who's superior based on number of patents.




--
※ 修改:·zamster 於 Jan 27 17:30:00 修改本文·[FROM: adsl-138-89-99-]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: adsl-138-89-99-]
发信人: zamster (彼岸花), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 17:29:55 2001) WWW-POST

I don't know Nasdaq runs on Tandem? Maybe you are right.
NYSE's trading system does run on Tandem, with backend
processing still on System 390. Now tandem is bought by
Compaq, how much future you think there is for that
platform?

【 在 dfehHo (ehHo) 的大作中提到: 】
: 【 在 devil (闲人) 的大作中提到: 】
: : http://www.sina.com.cn 2001/01/22 13:29 天极网
: :
: :   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: : 器临近崩溃的时候自动安全关闭服务器。
:
: Tandem nonstop server implemented it 20 years ago, that's
: why we never
: heard of Nasdaq got hacked/crashed.
:






--
※ 修改:·zamster 於 Jan 27 17:29:55 修改本文·[FROM: adsl-138-89-99-]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: adsl-138-89-99-]
发信人: laugh (haha), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 17:59:25 2001), 站内信件

ehey you don't have to defend IBM tthat much. :-PPP
OS2 was technically superior than win95?  it's a shame to compare the
superior OS/2 to win95! :-ppP

when i am retiring, i will consider IBM for my retirement. :))

【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: 【 在 pjones (彼得) 的大作中提到: 】
: : Concerning Nobel prize, I guess Lucent has more Nobel prizes than IBM does.
: : Does that prove anything? Not really.  Lucent is currently a big loser.
: You are right, past doesn't predict future
: : It doesn't make sense to honor a ping-pong champion a swimming title.
: : IBM is good at hardware development because they have tons of competitive
: : guys working on numerous projects along that direction.  However, their
: : software sucks---in every software field they lost to somebody: OS2 losts
: : to Windoz, AIX losts to Solaris and Linux, DB2 losts to Oracle.  What a
: : loser.  I happened to know some of their AI guys, don't see any brilliance.
: Concensus shows OS2 was technically superior than win95, DB2 beats Oracle
: as well at least on #transactions/s benchmark, IBM wasn't known for
: its marketing, but things will change as IBM starts agressive advertiseing.
: : If your smart colleague did something last night and you found your computer
: : acted weird this morning, guess you wouldn't feel good even if the colleague
: : claims he is trying to help.  If the computer happens to be a file server
: : and your boss just scolded you because he never ordered you to shutdown
: : the computer at 12:34am, then it is a more interesting scenario.
: The point is we are talking about smart AI that correctly detect system
: failure, most likely on server clusters in enterprise environment


--
  ,__o
_ \<,_
(*)/ (*)

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 216.192.196.38]
发信人: microbe (奇妙的微生物), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sat Jan 27 18:05:35 2001), 站内信件


【 在 laugh (haha) 的大作中提到: 】
: ehey you don't have to defend IBM tthat much. :-PPP
: OS2 was technically superior than win95?  it's a shame to compare the
: superior OS/2 to win95! :-ppP
: when i am retiring, i will consider IBM for my retirement. :))

and HP :-)

: 【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: : You are right, past doesn't predict future
: : Concensus shows OS2 was technically superior than win95, DB2 beats Oracle
: : as well at least on #transactions/s benchmark, IBM wasn't known for
: : its marketing, but things will change as IBM starts agressive advertiseing.
: : The point is we are talking about smart AI that correctly detect system
: : failure, most likely on server clusters in enterprise environment


--
我在门外坐了两天两夜,看着天空在不断的变化。
我才发现,虽然我到这里很久,却从来没有看清楚这片沙漠。

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 128.59.16.160]
发信人: pjones (彼得), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sun Jan 28 19:02:47 2001), 站内信件

【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: 【 在 pjones (彼得) 的大作中提到: 】
: : Concerning Nobel prize, I guess Lucent has more Nobel prizes than IBM does.
: : Does that prove anything? Not really.  Lucent is currently a big loser.
: You are right, past doesn't predict future
: : It doesn't make sense to honor a ping-pong champion a swimming title.
: : IBM is good at hardware development because they have tons of competitive
: : guys working on numerous projects along that direction.  However, their
: : software sucks---in every software field they lost to somebody: OS2 losts
: : to Windoz, AIX losts to Solaris and Linux, DB2 losts to Oracle.  What a
: : loser.  I happened to know some of their AI guys, don't see any brilliance.
: Concensus shows OS2 was technically superior than win95, DB2 beats Oracle
: as well at least on #transactions/s benchmark, IBM wasn't known for
: its marketing, but things will change as IBM starts agressive advertiseing.
Never believe those benchmark things which are quite bogus all the time.
Oracle/M$ SQLServer/IBM DB2/Informix all claim their products are the
fastest in the world.
However, the measurements were made on different hardware platforms with
different number of CPUs and different disk RAID products.  Nobody knows
what they are comparing. What I believe is there is no major difference
concerning performance among these 4 players.

Oracle's products are better in the sense they were the first to
support Programming Language extension (PL/SQL), spatial extension,
and time series extension.  DB2 has no PL extension and their
spatial extension is merely a blueprint paper right now.  DB2's progress
is even slower than M$'s SQLServer and Informix.  It can't compete with
Oracle on new markets like E-commerce.  Its market share comes from
mainframe because IBM is the only hardware manufacturer among the 4 major
players--only IBM can bind software with its hardware, which is a kind of
unfair competition as I believe.

: : If your smart colleague did something last night and you found your computer
: : acted weird this morning, guess you wouldn't feel good even if the colleague
: : claims he is trying to help.  If the computer happens to be a file server
: : and your boss just scolded you because he never ordered you to shutdown
: : the computer at 12:34am, then it is a more interesting scenario.
: The point is we are talking about smart AI that correctly detect system
: failure, most likely on server clusters in enterprise environment
There is a question haunting me from the beginning---what is the definition
of AI?  IMHO, a fanfare AI tech described in the advertisement which starts
all these arguments has never existed.  For the IBM AI guys, I guess
they have to face a dilemma to defend the fanfare:
On the one hand, if they have found a well-defined algorithm to detect
server crashing, then the result is an algorithm, not an artificial-
"intelligence" which will introduce unpredictability into system.
On the other hand, if they have found a probablistic algorithm (based
on Bayesian network or whatever) to detect server crashing, then the
result is un-predictable.  As a system admin I think this kind of thing
may be good for datamining, but not for system administration where
system must be recovered from any state to perfect(or seemingly perfect),
not from a less-inperfect state to an inperfect state.  Certainly I
assumed no program can recover a system as system recovery is a
philosophical problem harder than GO-game for the computers.
(At least GO-game has a well-defined goal---obtaining more crosspoints
on the board.  What is the well-defined goal for system recovery? I'd
like to see those AI guys' definition)

--
AND IF I TOLD YOU THAT I LOVED YOU,
YOU'D MAYBE THINK THERE'S SOMETHING WRONG.
I'M NOT A MAN OF TOO MANY FACES.
THE MASK I WEAR IS ONE.

※ 修改:.pjones 于 Jan 28 19:29:23 修改本文.[FROM: 164.67.26.150]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 164.67.26.150]
发信人: qili (qili), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Sun Jan 28 20:41:09 2001), 站内信件

【 在 pjones (彼得) 的大作中提到: 】
: : The point is we are talking about smart AI that correctly detect system
: : failure, most likely on server clusters in enterprise environment
: On the one hand, if they have found a well-defined algorithm to detect
: server crashing, then the result is an algorithm, not an artificial-
: "intelligence" which will introduce unpredictability into system.

I am not sure what distinguish an algorithm from AI. To me, AI is an attempt,
a decision making process carried out by machines to simulate what
an intelligent being like us would do in a similar environment.

for example, let's assume that most system admin would agree that if
a server is approaching 80% of its capacity, it should be rebooted. Well,
me, a device / algorithm that triggers a reboot at 80% of server capacity is
of AI.

: On the other hand, if they have found a probablistic algorithm (based
: on Bayesian network or whatever) to detect server crashing, then the
: result is un-predictable.  As a system admin I think this kind of thing

well, if the ill is unpredictable, the cure has to be equally unpredictable.
Untill we can write a 100% crash-proof server OS, we are going to get
random crashes now and then. Hopefully, before that happens, the server
has had its load taken over by others and get rebooted.

The whole discussion is kind of moot, as we don't know to what extent
the AI thing is to be applied. For example, in a single server, critical
application (an oxymoron itself), I would rather not have the AI thing.
In a redundant server farm where load can be rebalanced dynamically,
having this AI triggerman is pretty good.

--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.101.37.52]
发信人: magicfat (魔法胖子~missing my Friend), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 00:13:16 2001), 站内信件

【 在 qili (qili) 的大作中提到: 】
: well, if the ill is unpredictable, the cure has to be equally unpredictable.
: Untill we can write a 100% crash-proof server OS, we are going to get
: random crashes now and then. Hopefully, before that happens, the server
: has had its load taken over by others and get rebooted.
: The whole discussion is kind of moot, as we don't know to what extent
: the AI thing is to be applied. For example, in a single server, critical
: application (an oxymoron itself), I would rather not have the AI thing.
: In a redundant server farm where load can be rebalanced dynamically,
: having this AI triggerman is pretty good.

This debate got into some offense/defense on the ego of IBM(or IBMers)
so fast that my head couldn't stop spinning... 8-)
Seriously, I don't feel this new technology, though itself might
have some brilliant idea behind it, could help a lot in the particular
context of a running server.  The point is the service is interrupted,
no matter the server is restarted by a demon program, or a demon human
administrater.  Something that would make more sense is some better
load-balancing/failover technology.

--
他号令便号令好了,又何必安静?

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 38.204.0.2]
发信人: pjones (彼得), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 03:11:12 2001), 站内信件

【 在 qili (qili) 的大作中提到: 】
: 【 在 pjones (彼得) 的大作中提到: 】
: : On the one hand, if they have found a well-defined algorithm to detect
: : server crashing, then the result is an algorithm, not an artificial-
: : "intelligence" which will introduce unpredictability into system.
: I am not sure what distinguish an algorithm from AI. To me, AI is an attempt,
: a decision making process carried out by machines to simulate what
: an intelligent being like us would do in a similar environment.
: for example, let's assume that most system admin would agree that if
: a server is approaching 80% of its capacity, it should be rebooted. Well,
: me, a device / algorithm that triggers a reboot at 80% of server capacity is
: of AI.
hehe, in BBS talking nobody wants to explain entire ideas and nobody
wants to listen to other people's entire ideas either.  Sorry I made the
mistake and haven't explain my definition of AI.
IMHO, a deterministic algorithm is definitely not AI, otherwise everything
running by computer is AI which doesn't make sense.  My definition of
AI is those indeterministic algorithms with probablistic result, as
real "intelligence", this kind of algorithms sometimes are smart, and
sometimes make mistakes.  (According to this definition, game theory is
not AI, but machine learning based on Bayesian network is).  This
definition is idiosyncratic of mine, and I am sorry I didn't make it
clear at the very beginning (hehe, at that time I just thought I made
a short post).

Thus, if I was the designer of the reboot software, I would have done
something like this: if 95% swap area is used up, then reboot, blah
blah blah.  It's a deterministic algorithm progammable in any
procedural language.  Can't imagine what business AI has here.  It's
funny to think the system will reboot at 50% probability when 95%
swap area is used up.

: : On the other hand, if they have found a probablistic algorithm (based
: : on Bayesian network or whatever) to detect server crashing, then the
: : result is un-predictable.  As a system admin I think this kind of thing
: well, if the ill is unpredictable, the cure has to be equally unpredictable.
: Untill we can write a 100% crash-proof server OS, we are going to get
: random crashes now and then. Hopefully, before that happens, the server
: has had its load taken over by others and get rebooted.
: The whole discussion is kind of moot, as we don't know to what extent
: the AI thing is to be applied. For example, in a single server, critical
: application (an oxymoron itself), I would rather not have the AI thing.
: In a redundant server farm where load can be rebalanced dynamically,
: having this AI triggerman is pretty good.
hehe, here is the most disgusting thing I find in AI:  is it possible to
give a clear measurement that the AI tech makes difference here?  In other
words, will user loss anything if AI is not deployed here?  And to what
degree "this AI triggerman" is pretty good----I have seen too many buggy
programs and what if IBM gives us an "AI triggerman" program teemed with
bugs?  The answer is likely to be: you'll never know.  Well, then
a perfect AI program is not very different from a buggy AI program, then
why do we need an AI program at all?   As a programmer, my attitude
towards programming is very like Karl Popper's attitude towards science--
it must be refutable (verifiable).   Most of AI program lacks this feature
(some datamining AI program somewhat has this feature because visualization
of good clustering or bad clustering is easy to verify by human eyes) and I
totally can't see this feature in AI programs for fault-tolerance
applications.  Perhaps it is inappropriate to call those AI guys quackers,
but most of time they just fail to be persuasive.
--
AND IF I TOLD YOU THAT I LOVED YOU,
YOU'D MAYBE THINK THERE'S SOMETHING WRONG.
I'M NOT A MAN OF TOO MANY FACES.
THE MASK I WEAR IS ONE.

※ 修改:.pjones 于 Jan 29 03:17:33 修改本文.[FROM: 164.67.25.96]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 164.67.25.96]
发信人: qili (qili), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 14:31:30 2001), 站内信件

【 在 pjones (彼得) 的大作中提到: 】
: IMHO, a deterministic algorithm is definitely not AI, otherwise everything
: running by computer is AI which doesn't make sense.  My definition of
: AI is those indeterministic algorithms with probablistic result, as
: real "intelligence", this kind of algorithms sometimes are smart, and
: sometimes make mistakes.  (According to this definition, game theory is
: not AI, but machine learning based on Bayesian network is).  This
: definition is idiosyncratic of mine, and I am sorry I didn't make it
: clear at the very beginning (hehe, at that time I just thought I made
: a short post).

how do you make "indeterministic", or probablistic algorithms, and
implement them within a machine which by nature is deterministic?

short of introducing outside uncertainties, I don't see that
being done, even if you have a Baysian deduction mechnism in
your algorithms (Baysian deduction itself is deterministic,
not to mention that to convert a statistical probability into
a 0 / 1 decision itself has to be deterministic).

--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 206.189.24.10]
发信人: whh (大不了爱吃爱吃), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 16:16:55 2001), 站内信件

The state of a machine loaded with modern software is indeterministic.

【 在 qili (qili) 的大作中提到: 】
: 【 在 pjones (彼得) 的大作中提到: 】
: : IMHO, a deterministic algorithm is definitely not AI, otherwise everything
: : running by computer is AI which doesn't make sense.  My definition of
: : AI is those indeterministic algorithms with probablistic result, as
: : real "intelligence", this kind of algorithms sometimes are smart, and
: : sometimes make mistakes.  (According to this definition, game theory is
: : not AI, but machine learning based on Bayesian network is).  This
: : definition is idiosyncratic of mine, and I am sorry I didn't make it
: : clear at the very beginning (hehe, at that time I just thought I made
: : a short post).
: how do you make "indeterministic", or probablistic algorithms, and
: implement them within a machine which by nature is deterministic?
: short of introducing outside uncertainties, I don't see that
: being done, even if you have a Baysian deduction mechnism in
: your algorithms (Baysian deduction itself is deterministic,
: not to mention that to convert a statistical probability into
: a 0 / 1 decision itself has to be deterministic).


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 128.107.148.80]
发信人: pjones (彼得), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 17:59:59 2001), 站内信件

【 在 qili (qili) 的大作中提到: 】
: 【 在 pjones (彼得) 的大作中提到: 】
: : IMHO, a deterministic algorithm is definitely not AI, otherwise everything
: : running by computer is AI which doesn't make sense.  My definition of
: : AI is those indeterministic algorithms with probablistic result, as
: : real "intelligence", this kind of algorithms sometimes are smart, and
: : sometimes make mistakes.  (According to this definition, game theory is
: : not AI, but machine learning based on Bayesian network is).  This
: : definition is idiosyncratic of mine, and I am sorry I didn't make it
: : clear at the very beginning (hehe, at that time I just thought I made
: : a short post).
: how do you make "indeterministic", or probablistic algorithms, and
: implement them within a machine which by nature is deterministic?
: short of introducing outside uncertainties, I don't see that
: being done, even if you have a Baysian deduction mechnism in
: your algorithms (Baysian deduction itself is deterministic,
: not to mention that to convert a statistical probability into
: a 0 / 1 decision itself has to be deterministic).

Yes, a non-interactive program is always deterministic.  However,
you can introduce some entropy from user/network into the program as
parameters (i.e. "outside uncertainty")
For example, random() system call is always pseudo-random if seed
is given.  You may let user choose seed on the fly, hence the
genuine randomness and indeterminism.
For those servers, workload introduced by users' behavior is considered
to be totally random.
--
AND IF I TOLD YOU THAT I LOVED YOU,
YOU'D MAYBE THINK THERE'S SOMETHING WRONG.
I'M NOT A MAN OF TOO MANY FACES.
THE MASK I WEAR IS ONE.

※ 修改:.pjones 于 Jan 29 18:06:01 修改本文.[FROM: 131.179.64.111]
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 131.179.64.111]
发信人: qili (qili), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 20:01:51 2001), 站内信件

【 在 pjones (彼得) 的大作中提到: 】
: Yes, a non-interactive program is always deterministic.  However,
: you can introduce some entropy from user/network into the program as
: parameters (i.e. "outside uncertainty")
: For example, random() system call is always pseudo-random if seed
: is given.  You may let user choose seed on the fly, hence the
: genuine randomness and indeterminism.

hold a second. is there a difference between an indeterministic
or probablistic algorithm and a deterministic algorithm with
a random input?

For example, the algorithm in random() is deterministic: for the same
seed, it gives you precisely the same sequence everytime. This is
a deterministic algorithm with a random input (the seed).

On the other hand, noise out of a zehner diode is random, in the
sense that the noise is unpredictable due to the mechanism
by which the noise is generated.

to me, the two are fundamentally different.

--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.101.37.70]
发信人: dudu (神奇主), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 20:55:11 2001) WWW-POST

真是一篇好文章。尤其是对AI的分析。
记得在国内上AI课时,老师讲起一个著名的笑话:
AI的目标就是摘月亮。一天有一个AIer说他有重大进展。
大家一看,原来他站到房顶上去了。他说可以证明在房顶上
更容易摘到月亮。
【 在 pjones (彼得) 的大作中提到: 】
: 【 在 weijing (味精号巡洋舰) 的大作中提到: 】
: : 【 在 pjones (彼得) 的大作中提到: 】
: : : Concerning Nobel prize, I guess Lucent has more Nobel prizes than IBM does.
: : : Does that prove anything? Not really.  Lucent is currently a big loser.
: : You are right, past doesn't predict future
: : : It doesn't make sense to honor a ping-pong champion a swimming title.
: : : IBM is good at hardware development because they have tons of competitive
: : : guys working on numerous projects along that direction.  However, their
: : : software sucks---in every software field they lost to somebody: OS2 losts
: : : to Windoz, AIX losts to Solaris and Linux, DB2 losts to Oracle.  What a
: : : loser.  I happened to know some of their AI guys, don't see any brilliance.
: : Concensus shows OS2 was technically superior than win95, DB2 beats Oracle
: : as well at least on #transactions/s benchmark, IBM wasn't known for
: : its marketing, but things will change as IBM starts agressive advertiseing.
: Never believe those benchmark things which are quite bogus all the time.
: Oracle/M$ SQLServer/IBM DB2/Informix all claim their products are the
: fastest in the world.
: However, the measurements were made on different hardware platforms with
: different number of CPUs and different disk RAID products.  Nobody knows
: what they are comparing. What I believe is there is no major difference
: concerning performance among these 4 players.
:
: Oracle's products are better in the sense they were the first to
: support Programming Language extension (PL/SQL), spatial extension,
: and time series extension.  DB2 has no PL extension and their
: spatial extension is merely a blueprint paper right now.  DB2's progress
: is even slower than M$'s SQLServer and Informix.  It can't compete with
: Oracle on new markets like E-commerce.  Its market share comes from
: mainframe because IBM is the only hardware manufacturer among the 4 major
: players--only IBM can bind software with its hardware, which is a kind of
: unfair competition as I believe.
:
: : : If your smart colleague did something last night and you found your computer
: : : acted weird this morning, guess you wouldn't feel good even if the colleague
: : : claims he is trying to help.  If the computer happens to be a file server
: : : and your boss just scolded you because he never ordered you to shutdown
: : : the computer at 12:34am, then it is a more interesting scenario.
: : The point is we are talking about smart AI that correctly detect system
: : failure, most likely on server clusters in enterprise environment
: There is a question haunting me from the beginning---what is the definition
: of AI?  IMHO, a fanfare AI tech described in the advertisement which starts
: all these arguments has never existed.  For the IBM AI guys, I guess
: they have to face a dilemma to defend the fanfare:
: On the one hand, if they have found a well-defined algorithm to detect
: server crashing, then the result is an algorithm, not an artificial-
: "intelligence" which will introduce unpredictability into system.
: On the other hand, if they have found a probablistic algorithm (based
: on Bayesian network or whatever) to detect server crashing, then the
: result is un-predictable.  As a system admin I think this kind of thing
: may be good for datamining, but not for system administration where
: system must be recovered from any state to perfect(or seemingly perfect),
: not from a less-inperfect state to an inperfect state.  Certainly I
: assumed no program can recover a system as system recovery is a
: philosophical problem harder than GO-game for the computers.
: (At least GO-game has a well-defined goal---obtaining more crosspoints
:  on the board.  What is the well-defined goal for system recovery? I'd
:  like to see those AI guys' definition)
:


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: rb16v-93-229.re]
发信人: whh (大不了爱吃爱吃), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 20:58:47 2001), 站内信件

What if I write a random() algorithm that takes the input of quantified
value of noise out of a zehner diode as the random number it returns.
The algorithm, according to you, is deterministic, the output is
totally unpredictable. What is the fundamental difference between
these two methods to generate randomness?

【 在 qili (qili) 的大作中提到: 】
: 【 在 pjones (彼得) 的大作中提到: 】
: : Yes, a non-interactive program is always deterministic.  However,
: : you can introduce some entropy from user/network into the program as
: : parameters (i.e. "outside uncertainty")
: : For example, random() system call is always pseudo-random if seed
: : is given.  You may let user choose seed on the fly, hence the
: : genuine randomness and indeterminism.
: hold a second. is there a difference between an indeterministic
: or probablistic algorithm and a deterministic algorithm with
: a random input?
: For example, the algorithm in random() is deterministic: for the same
: seed, it gives you precisely the same sequence everytime. This is
: a deterministic algorithm with a random input (the seed).
: On the other hand, noise out of a zehner diode is random, in the
: sense that the noise is unpredictable due to the mechanism
: by which the noise is generated.
: to me, the two are fundamentally different.


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 128.107.148.80]
发信人: qili (qili), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Mon Jan 29 21:18:36 2001), 站内信件

【 在 whh (大不了爱吃爱吃) 的大作中提到: 】
: What if I write a random() algorithm that takes the input of quantified
: value of noise out of a zehner diode as the random number it returns.
: The algorithm, according to you, is deterministic, the output is
: totally unpredictable. What is the fundamental difference between
: these two methods to generate randomness?

well, the answer depends on your perspective.

A little earlier, I stated that, as an hypothetic example,
a deterministic rule can behave like a human, thus giving it
"intelligence". A counter-argument was made that such a deterministic
algorithm cannot be AI. Upon further examination, I thought
the essence of that argument was on the indeterministic
nature of the output of that algorithm, not the algorithm
itself (see pjones' example of random() discussion).

Going back to your example, the simplest algorithm would be to
just pass on the random value obtained from the diode. if you
define "algorithm" here as the diode plus the "passing on" part,
then you have an indeterministic algorithm; if you define
"algorithm" narrowly as the "passing on" part only,
you have a deterministic algorithm. In the later example,
a deterministic value, together with a random input (the diode).

Such an example further strengthens the argument that
being AI or not, at least within the framework of pjones, does
not require the algorithm itself be indeterministic.



--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 32.101.37.220]
发信人: Antz (四海为家), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Thu Feb  1 00:12:40 2001), 转信

呵呵,我们AI课的第一张幻灯片上只有一句话:
AI is something taht can beat most human stupidity.

【 在 dudu (神奇主) 的大作中提到: 】
: 真是一篇好文章。尤其是对AI的分析。
: 记得在国内上AI课时,老师讲起一个著名的笑话:
: AI的目标就是摘月亮。一天有一个AIer说他有重大进展。
: 大家一看,原来他站到房顶上去了。他说可以证明在房顶上
: 更容易摘到月亮。
: 【 在 pjones (彼得) 的大作中提到: 】
: : Never believe those benchmark things which are quite bogus all the time.
: : Oracle/M$ SQLServer/IBM DB2/Informix all claim their products are the
: : fastest in the world.
: : However, the measurements were made on different hardware platforms with
: : different number of CPUs and different disk RAID products.  Nobody knows
: : what they are comparing. What I believe is there is no major difference
: : concerning performance among these 4 players.
: : Oracle's products are better in the sense they were the first to
: : support Programming Language extension (PL/SQL), spatial extension,
: : and time series extension.  DB2 has no PL extension and their
: : spatial extension is merely a blueprint paper right now.  DB2's progress
: : is even slower than M$'s SQLServer and Informix.  It can't compete with
: : Oracle on new markets like E-commerce.  Its market share comes from
: : mainframe because IBM is the only hardware manufacturer among the 4 major
: : players--only IBM can bind software with its hardware, which is a kind of
: : unfair competition as I believe.
: : There is a question haunting me from the beginning---what is the definition
: : of AI?  IMHO, a fanfare AI tech described in the advertisement which starts
: : all these arguments has never existed.  For the IBM AI guys, I guess
: : they have to face a dilemma to defend the fanfare:
: : On the one hand, if they have found a well-defined algorithm to detect
: : server crashing, then the result is an algorithm, not an artificial-
: : "intelligence" which will introduce unpredictability into system.
: : On the other hand, if they have found a probablistic algorithm (based
: : on Bayesian network or whatever) to detect server crashing, then the
: : result is un-predictable.  As a system admin I think this kind of thing
: : may be good for datamining, but not for system administration where
: : system must be recovered from any state to perfect(or seemingly perfect),
: : not from a less-inperfect state to an inperfect state.  Certainly I
: : assumed no program can recover a system as system recovery is a
: : philosophical problem harder than GO-game for the computers.
: : (At least GO-game has a well-defined goal---obtaining more crosspoints
: :  on the board.  What is the well-defined goal for system recovery? I'd
: :  like to see those AI guys' definition)


--

    君子无爵而贵,无禄而富,不言而信,不怒而威,处穷而荣,独居而乐.
岂不至尊,至富,至重,至严之情举积此哉?
                                                --荀子

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 128.249.124.52]
发信人: ayanami (我是可爱的男甲亢), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Thu Feb  1 10:13:40 2001), 转信

hehe mark, mark.

【 在 Antz (四海为家) 的大作中提到: 】
: 呵呵,我们AI课的第一张幻灯片上只有一句话:
: AI is something taht can beat most human stupidity.
: 【 在 dudu (神奇主) 的大作中提到: 】
: : 真是一篇好文章。尤其是对AI的分析。
: : 记得在国内上AI课时,老师讲起一个著名的笑话:
: : AI的目标就是摘月亮。一天有一个AIer说他有重大进展。
: : 大家一看,原来他站到房顶上去了。他说可以证明在房顶上
: : 更容易摘到月亮。


--
The metric system did not really catch on in the States, unless you
count the increasing popularity of the nine-millimeter bullet.

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 64.32.197.50]
发信人: microbe (奇妙的微生物), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Thu Feb  1 23:05:54 2001), 站内信件


Where is the original annoucement?  Can't find on IBM website.

【 在 devil (闲人) 的大作中提到: 】
: http://www.sina.com.cn 2001/01/22 13:29 天极网
:   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: 器临近崩溃的时候自动安全关闭服务器。
:   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: 全关机,防患于蔚然。
:   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: 术称做“软件再生”技术。(竹子编译)


--
我在门外坐了两天两夜,看着天空在不断的变化。
我才发现,虽然我到这里很久,却从来没有看清楚这片沙漠。

※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: 160.39.200.28]
发信人: dfehHo (ehHo), 信区: ITnews
标  题: Re: IBM开发出预测服务器崩溃的软件
发信站: The unknown SPACE (Fri Feb  2 22:44:14 2001) WWW-POST


http://news.cnet.com/news/0-1003-200-4550757.html

【 在 microbe (奇妙的微生物) 的大作中提到: 】
:
: Where is the original annoucement?  Can't find on IBM website.
:
: 【 在 devil (闲人) 的大作中提到: 】
: : http://www.sina.com.cn 2001/01/22 13:29 天极网
: :   世界上最大的计算机制造商IBM称,它已经研究开发出一种软件,可以在服务
: : 器临近崩溃的时候自动安全关闭服务器。
: :   IBM低端XSeries系列服务器结构技术研究负责人Tom Bradicich称,现在的服
: : 务器可以正常关机,重启,避免任何软件被死锁,但是它们都还没有预测服务器崩
: : 溃时间的功能。而新开发的这套软件采用了人工智能,可以在发生意外之前自动安
: : 全关机,防患于蔚然。
: :   目前还有多家公司在开发类似的软件,以减少服务器当机造成的巨大损失。例
: : 如Sun微系统公司,它们正在销售的一种软件可以对服务器的问题做出反应,或自
: : 动到网上下载升级软件。IBM认为他们在这方面还处于领先地位,他们将自己的技
: : 术称做“软件再生”技术。(竹子编译)
:
:


--
※ 来源:.The unknown SPACE bbs.mit.edu.[FROM: ]

[返回]
赞助链接
未名交友
将您的链接放在这儿
 

Site Map - Contact Us - Terms and Conditions - Privacy Policy

版权所有,未名空间(mitbbs.com),since 1996