From: Frye, Matthew (Matthew.Frye@REXHEALTH.COM)
Date: Fri Jan 16 2004 - 09:24:17 EST
I/O bandwidth and data loaded in memory _are_ important considerations, but
the point I made was based on my experience, not on what is "likely," or
what is true in theory. However, and additional caveat...when I said users,
mostly I was referring to DBAs. End users may not notice (or complain), but
DBAs certainly seem to notice, if they are doing their job. This also
varies based on how much they are paying attention and relative whininess.
In my experience, SAP DBAs are more whiny than other DBAs.
Matt Frye
Sr. Systems Programmer, RS/6000 Group
Rex Healthcare
(919) 784-3791
> ----------
> From: Green, Simon[SMTP:Simon.Green@EU.ALTRIA.COM]
> Reply To: IBM AIX Discussion List
> Sent: Thursday, January 15, 2004 12:03 PM
> To: aix-l@Princeton.EDU
> Subject: Re: Migratepv Command
>
> Whilst it's true that migratepv can have a serious impact on I/O, I've
> used
> it during normal working hours on many occasions without anybody
> complaining. I've even moved multiple LVs in parallel without problems,
> (but I've only done this where the disks were using different hardware:
> different adapters, on different I/O busses).
>
> I would say that the most important factor is I/O bandwidth. If there's
> sufficient capacity then users probably won't notice anything.
>
> Memory is also important. If the data that the users want is already
> loaded
> into memory then it doesn't really matter what you do on the disks,
> provided
> there's enough CPU capacity to service the I/O requests. If you're doing
> a
> lot of paging, you're more likely to see performance degradation although
> even then, if the data is in paging space it doesn't matter what happens
> to
> the actual filesystems. Of course, the data that the users are most
> likely
> to want should be the stuff that's in memory, so in theory it's unlikely
> that they'll want to load any pages from the LV you're moving.
>
> It IS dependent on the particular workload as well as the I/O
> infrastructure, but although caution is appropriate I wouldn't necessarily
> anticipate disastrous results.
>
> For someone unfamiliar with using migratepv, start off with a smallish LV
> during a period of low activity and see what happens, then work up.
>
> --
> Simon Green
> Altria ITSC Europe Ltd
>
> AIX-L Archive at https://new-lists.princeton.edu/listserv/aix-l.html
>
> New to AIX? http://publib-b.boulder.ibm.com/redbooks.nsf/portals/UNIX
>
> N.B. Unsolicited email from vendors will not be appreciated.
> Please post all follow-ups to the list.
>
> > -----Original Message-----
> > From: Frye, Matthew [mailto:Matthew.Frye@REXHEALTH.COM]
> > Sent: 15 January 2004 16:33
> > To: aix-l@Princeton.EDU
> > Subject: Re: Migratepv Command
> >
> >
> > Oracle doesn't know about it, but you can bet the users will notice.
> > Migratepv -l has been known to have a noticeable effect on
> > I/O, especially
> > with databases. My suggestion would be to coordinate a time
> > when the system
> > usage is low(ish). Of course, it depends on the application.
> > That time may
> > not exist for your environment.
>
This archive was generated by hypermail 2.1.7 : Wed Apr 09 2008 - 22:17:31 EDT