Sabi ni fooler noong Tue, 02 May 2000 13:09:43 +0800
> the main reason is still growing, lots of updates in the kernel area and
still need
> to reboot for that updates to take effect to improve reliability.
But I thought that some kernel updates didn't require you to reboot? If so,
then it's really something inherent in the kernel that causes reboots. I
remember I threads on this list -- or I could'a been dreaming ;-) -- about how a
buggy display driver can hang X, and take down Linux. In fact, from my
recollection of the discussions here, it's mainly on X that Linux hangs, so it's
X's fault. Still, it shouldn't be able to take down the entire OS if the kernel
were robust enough to begin with.
> its really nice that you have a box that doesnt reboot since from its startup
but
> unfortunately in linux world, there is what we called a memory leak problem.
memory
> leak cause by a certain application that doesnt clean it up well after
allocating
> some portions of memory. this will leave a black hole in your memory area.
even upto
> date linux kernel has still a memory leak problem (and also NT server but i
dont
> know in mainframes but i believe there are also some memory leak problems
too).
Not having easy access to a mainframe, I can't say whether they do have memory
leak problems. I suspect, however, that they don't because of the robust memory
protection of a mainframe OS. I know that x86 programmers bitch about its
segmented memory architecture, but IMHO, it's gotten a bad rap, and actually
makes memory protection easier from a systems programming viewpoint. Apps
programmers might not see that. Some mainframes do have segmented memory -- some
models of the Borroughs series, now owned by Unisys, come to mind. Segmenting
memory makes it easier to tag an entire segment of memory as belonging to a
process, and no other process can touch it. Even for flat-space processors,
however, such as IBM's POWER2, the OS designers went to great lengths to ensure
that memory leaks would not take place.
The other thing about mainframes is that they do crash, but recover quickly
enough so that the users don't notice. I remember "Crash-Proof Computing", an
excellent article by Tom Halfhill in Byte magazine, which describes how
reliability and availability are the main goals of mainframe hardware and OS
designers -- not eye-or-ear candy. The other thing is, a mainframe's hardware
and software are more tightly tied together, in contrast with a PC, where there
are literally millions of different hardware configurations.
> regarding to mainframes that doesnt reboot since startup, the same true to
other
> operating system if you already contended with your setup. why upgrade and
reboot if
> you already contended with your setup right?
Yes. Which points to the maturity of big-iron OSes compared to Linux. Will
Linux ever get there? I think it has the potential, by improving its memory
management and protection. Heck, most of it is done in hardware, starting from
the 386. Question is whether kernel changes in that direction will break older
apps.
Going back to the "dream machine" thread, my own concept is a dual-box system
-- one running nothing but X, and the other serving apps, and the two connected
via Ethernet.
Mabuhay.
-
Philippine Linux Users Group. Web site and archives at http://plug.linux.org.ph
To leave: send "unsubscribe" in the body to [EMAIL PROTECTED]
