On May 13, 2005, at 12:05 PM, Brian Chee wrote:
Actually I have a question...why would you want to run a machine
without
swap? There are good reasons if you're running an embedded linux
machine,
but for normal machines I've seen folks setup unix boxes that boot
from the
network but ONLY do swap to the local hard disk. Those old xterms did
this
alot just so that you don't swap over the network.
A good example of a linux machine that didn't use swap were those old
N!C
(New Internet Computers) that Larry Ellison of Oracle pushed for a
while.
A somewhat better and more recent example is a cluster. Cluster
applications that process large amounts of data, such as parallel
scientific or multimedia applications, are likely to cause swapping on
individual cluster nodes, for reasons we've already discussed on-list.
These applications will perform better on clusters with network
swapping support.
Network swapping allows any cluster node with over-committed memory to
use idle memory of a remote node as its backing store and to “swap” its
pages over the network. As the disparity between network speeds and
disk speeds continues to grow, network swapping will be faster than
traditional swapping to local disk.
A UDMA-133 disk has a theoretic throughput of 133MBps, or just a tad
over 1Gbps, which is fairly cheap Ethernet gear these days. Course,
your HDD won't actually supply data at 1Gbps, while zero-copy
networking is a reality on linux, and thus, it might actually be faster
these days to hit the wire or fiber rather than the rotating glass
covered in oxide.
These ran from the internal cdrom and had no swap...while a good
concept,
they were pigs...very slow and they were cheap only for a very short
amount
of time. Once you tossed in a small laptop drive for swap, boy did
they run
great...but once you did that you lost the cost advantage.
The only embedded machines I've seen nowdays that regularlly don't run
any
swap at all, are those tiny little CPM or DOS machines.
Take a look in your lab Brian. Does any of that gear run linux without
a HDD?
Since those don't even have virtual memory, swap isn't necessary.
Once again, swap != virtual memory. (paging == virtual memory!!) Way
back in 1970 or so
you could run unix on a PDP-7 with no virtual memory support. (Hack,
even 32/V from AT&T didn't support virtual memory on the Vax!). You
could (and likely *would*), swap, however.
UC Berkley managed to make a VM system for (BSD) Unix, and the world
changed. :-) Course, putting TCP/IP in there might have had something
to do with it too.
But on a multithreaded/multiuser machine swap just makes sense.
Only to serve as a place to stuff otherwise cold dirty pages. Disk is
*so* slow compared to memory these days that even the database people
like to keep the DB in-core, if at all possible.
Anyway, forgive me if I missed the point...but at the super low cost of
drives nowdays, even a tiny drive for swap makes sense.
My disk drives cost more than my 802.11 APs, which run linux, and don't
have a swap area. What am I doing wrong?
jim
