Date: Wed, 21 Mar 2001 13:43:44 -0800 From: Mike Smith <msmith@freebsd.org> To: Poul-Henning Kamp <phk@critter.freebsd.dk> Cc: arch@freebsd.org Subject: Re: remind me again, why is MAXPHYS only 128k ? Message-ID: <200103212143.f2LLhih02089@mass.dis.org> In-Reply-To: Your message of "Wed, 21 Mar 2001 22:17:21 %2B0100." <89046.985209441@critter>
next in thread | previous in thread | raw e-mail | index | archive | help
> In message <200103212111.f2LLBSh01790@mass.dis.org>, Mike Smith writes:
> >>
> >> Are there any roadblocks for increasing MAXPHYS as a tweakable these
> >> days, or is it still an "do not alter or ELSE..." #define ?
> >
> >There isn't much hardware out there that can do anything useful with an
> >I/O larger than 128k.
>
> Well, while that is true, things like striping and raid-5 could benefit
> from not being limited to a 128k stripesize for certain kinds of
> bulk application.
You're referring here to the interface between the bio layer and the RAID
layer, rather than the actual limit on a physical I/O though. I realise
that the RAID layer in question pretends to be an I/O layer, hence this
being an issue.
I think that MAXPHYS should become DFLTMAXPHYS, and the si_iosize_max
field should be allowed to be an override in either direction, so that
devices that *are* OK with larger transfers can actually take advantage
of them.
> I also belive tapes have been mentioned as a candidate for a larger
> MAXPHYS.
This I think has mostly been the case with SGI systems and large tape
blocks, I think, but I don't recall whether 128k was enough.
> >There was also a lengthy discussion on I/O
> >saturation a little while back; the short answer is just that making it
> >larger doesn't win anything significant, and may cause unacceptable
> >latencies in some cases.
>
> Right, I'm not asking for a general increase, I'm asking if it can
> be increased on a case-by-case basis or if the system will explode
> for arcane architectural reasons if I double or quadruple it on
> a particular system ?
I don't believe that any of the previous discussion uncovered anything
that would explode. You would probably want to check though. 8)
--
... every activity meets with opposition, everyone who acts has his
rivals and unfortunately opponents also. But not because people want
to be opponents, rather because the tasks and relationships force
people to take different points of view. [Dr. Fritz Todt]
V I C T O R Y N O T V E N G E A N C E
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200103212143.f2LLhih02089>