BBS水木清华站∶精华区

发信人: nudtbegger (军人), 信区: Linux
标  题: Re: 再论"对linux的质量质疑"
发信站: BBS 水木清华站 (Sat Jul 31 09:05:14 1999) WWW-POST

【在 fzhang (老铁) 的大作中提到: 】
∶ Yes, it is true in Linux2.2, but will not in 2.4(we can expect it this
∶ fall):
∶ The Linux model of network sockets is one common across most UNIX variants
∶
∶ which adhere to the standards. This is not to say that the standard doesn't
∶
∶ have some deficiencies. Under Linux 2.2, if you have a number of processes
∶
∶ all waiting on an event from a network socket (a web port, for instance)
∶ they will all be woken up when activity is detected. So, for every web page
∶
∶ request, a number of web server processes will suddenly wake up and try and
∶
∶ get at the request... only one will succeed. The losers of the herd will
∶ put
∶ themselves back to sleep and wait for the next connection.  Linux is a
∶ terribly
∶ efficient system and manages to wake up all these processes, give the
∶ resource
∶ to one of them, and put the remainder back to sleep very quickly, however
∶ benchmarks have shown that we could be more efficient if we eliminated the
∶
∶ "stampede effect." Thus, the Linux minds cast about for a solution and
∶ decided
∶  to implement "wake one" under Linux. This change allows programs to let
∶ Linux
∶ know that they should be the only one to be woken up in the case of
∶ activity.
∶ It is expected that by the time Linux 2.4 ships, Apache and other similar
∶ se
∶ rvers will begin using this new functionality to increase the efficiency of
∶
∶ what they do.
老兄，你这段英文为什么不翻成汉语呢？看的我好费劲。
2.4里都加了些啥呢？what 's new function?

∶ 中断吗?那是异步驱动的,Linux不也可以吗?
∶ 但操作系统可以使得用户程序具有阻塞语义的
what's mean?

∶ ?: 真有要求的话,Linus推出的会更快的
∶ 这倒是,操作系统变化挺快的,但NT太笨重了,跟不上吧,NT2000怕要难产

--
※ 来源:·BBS 水木清华站 bbs.net.tsinghua.edu.cn·[FROM: 202.197.0.1]