On Sat, 3 Jun 2000 lists@frednet.dyndns.org wrote:
>
> Don't the HP RISC line have a feature kind of like that? I believe
> they are able to load a new kernel without a reboot or even downtime.
> IMHO, this sounds like a great idea, and would definitely be a "plus"
> for public relation encouragement. Also, the same case with the Hot
> Swappable CPU idea, this would be great for articles and reviews,
> (never mind the obvious practical benefits).
There are quite a few specialist systems which do similar things; the one
which springs to mind is Tandem's Himalaya series, but they aren't the
only player by any means.
Right now, it's a very specialised area - but it is certainly becoming
less so. Servers like Sun's E10k Starfire systems implement partitioning,
for example, and things like hotplug PCI are appearing. Soon, this sort of
N+N redundancy could be relatively common in high-end servers: it would be
nice to see Linux getting there first...
> To the man that just wanted to add another box instead: Just think of
> it this way, nearly everything that gets done in kernel development
> and added features, there is probably a better way to do it with
> another solution, but the ability to do it is what makes the
> difference. Look at Microsoft: supposedly, you can run a box with up
> to 32 CPUs with win2k maximus serveros(or whatever), but you could
> probably do a much better job with 32 seperate machines and proper
> load balancing, but 32 concurrent CPUs sure made them look nice in
> reviews.
For some things, yes, 32 individual machines are as good as one machine
with 32x the power. There are plenty of tasks where the extra overhead of
clustering machines would kill the performance, though: only certain
categories of number-crunching split up well enough to throw a Beowulf
cluster at the job.
> It's stuff like that (the nonsense, stupid things) that make
> the difference. Also, I sure wouldn't mind having the capacity to
> take out the second CPU on my box if one fails, or I just want to (I
> don't know the practicalities of this) upgrade it for that matter, but
> I'm sure big businesses using boxes like that sure would eat that
> feature like candy. It might even contributed to linux's reliable
> reputation. Just my .02.
There are applications where having a pair of boxes, one of which is
always available, just isn't good enough: you need ONE box, available
24x7. In these areas, the customer is more than willing to hand over $$$
for an extra 9 of availability. If the bank's server b0rks half-way
through a transaction, of course you can get the data back once the
machine reboots and fscks, then processes all the transaction logs. This
just isn't good enough, though: it needs to work first time, every time.
James.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sat Jun 03 2000 - 16:04:28 EDT