[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [rhn-users] HELP!! Bad blocks when installing RH9. What should I do?

>>>>> "Hattie" == Hattie Rouge <redhat netgods us> writes:

Hattie> See below...  Hattie Rouge

>> -----Original Message----- From: rhn-users-admin redhat com
>> Edward C. Bailey Sent:
>> Already is -- you don't *have* to have swap -- but if you need it,
>> you'll wish you had it... :-)

Hattie> Oh, this is good news.  I hadn't run across this tidbit.  When did
Hattie> it show up, do you know?

As far as I know it's always been the case -- if the system doesn't do a
swapon(), it won't be using swap... :-)

Hattie> Anyone know if you can delete swap on the fly?  That's another nice
Hattie> thing about Solaris.  Even on a production system you can safely
Hattie> delete swap and then if you don't have enough RAM, add it back.
Hattie> And yes, I've done it on a production Oracle server.  I had enough
Hattie> RAM for almost all of the workload.

To be honest, I don't know; doing a swapoff changes what free displays, but
that doesn't mean that I'd feel comfortable pull swap out from underneath a
production system.

And to be honest, why would you?  On a system with a properly
designed/configured virtual memory subsystem, it's not like the mere
existence of swap is going to affect system performance...

>> It's already been done -- IBM was building drives with two separate
>> access arms (and control/data paths) for its mainframes about 20 years
>> ago.  If I recall correctly, they weren't used for splitting OS/swap I/O
>> loads, but to support multi-system access to the drives.  I don't know
>> whether they still do this (though I'd guess not, given how there are
>> other ways to solve this particular problem)...

Hattie> I meant with modern drives.  I would think that a significant PC
Hattie> fraction would pay a bit more to get the extra I/O performance if
Hattie> the performance is significant.  That's a lot of disks...

The problem is that the price wouldn't be "a bit more" -- you'd need two
head/arm/actuator assemblies, two sets of controller electronics, and a
modified drive baseplate.  The costs alone for these components would
likely get you to nearly double the cost of a single-arm drive, and the
fact that the resulting product would be a low-volume, specialty item, and
you'll easily be paying more than twice the cost of a single-arm drive.

The PC hardware industry is so competitive, if there truly was an advantage
to something like this, it would be on the market already.  Heck, this kind
of hardware would be *perfect* for a Tivo-like environment, but apparently
it's cheaper to just use commodity hardware...

The fact of the matter is that it's just as effective to just get two
drives, and dedicate one to swap.  True, it's not as elegant, but elegance
doesn't make systems run any better -- brute force solutions can be just as
effective... :-)

Ed Bailey        Red Hat, Inc.          http://www.redhat.com/

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]