So IOPS is not the right way of measuring for mainframe data

So there is a big difference between IOPS and Response time

On Sun, 8 Sept 2024, 16:37 Nigel Morton, <
[email protected]> wrote:

> IOPS means I/Os per second so it's a measure of the throughput of the
> devices. In practice, with z/OS you won't generate anywhere near the
> maximum IOPs for a modern all-flash storage device. IMHO response time is a
> more useful measure for performance especially as IOPs figures will
> probably come from synthetic benchmarks.
>
> Also consider how you'll migrate from an old storage system to a new one.
> Copy services functions such as multi-target metro mirror are excellent but
> will almost certainly require that the source and target vendor are the
> same. You might also need to consider what type of FICON adapters can
> coexist with your FICON switches and channels.
>
> On Sun, 8 Sept 2024 at 05:04, Peter <
> [email protected]> wrote:
>
> > Hello
> >
> >
> > I have seen a term called IOPS in the technical sheet
> >
> >
> >
> > From the storage perspective what does IOPS do? How does that influence
> on
> > storage solution selection?
> >
> > On Thu, 5 Sept 2024, 16:03 Radoslaw Skorupka, <
> > [email protected]> wrote:
> >
> > > Cache - the more the better. Obvious. However there are different ways
> > > of use of DRAM as a cache.
> > > Some vendors have cache mirrored. Others don't mirror READ cache.
> > > Usually the same memory is used for other purposes, not only for
> caching
> > > READs and WRITEs. So, "1TB of cache" is not simply comparable between
> > > vendors.
> > >
> > > Disk module. SSD or magnetic? Usually SSD nowadays. But which SSD?
> There
> > > are many flavours and capacities.
> > > Usually smaller modules gives better overall performance. And better
> > > flexibility. And there are faster and slower modules.
> > >
> > > Remote copy.
> > > Do you want to perform PiT copy on secondary/slave?
> > > Do you want asynchronous copy?
> > > Do you want PPRC and FLASHCOPY command compatibility?
> > > What QoS features of PiT and Remote Copy do you want?
> > > etc.
> > >
> > >
> > > --
> > > Radoslaw Skorupka
> > > Lodz, Poland
> > >
> > >
> > >
> > >
> > > W dniu 04.09.2024 o 20:27, Peter pisze:
> > > > Thanks for the inputs
> > > >
> > > > Currently we are metro mirrored.
> > > >
> > > > So I believe I also need to look into their quality of support?
> > > >
> > > > Cache and modules? -  what exactly I need to be researching on this ?
> > > >
> > > > On Wed, 4 Sept 2024, 19:40 Radoslaw Skorupka, <
> > > > [email protected]> wrote:
> > > >
> > > >> W dniu 04.09.2024 o 10:52, Peter pisze:
> > > >>> Hello,
> > > >>>
> > > >>> I Must confess that I am not a Storage expert.  In our organization
> > we
> > > >> are
> > > >>> working on procuring a new storage box solution. I have been asked
> to
> > > >>> study and give an opinion on which solution would be perfect for
> our
> > > >>> mainframe environment.
> > > >>>
> > > >>> Could someone give me your advise based on your experience. When
> > > >> selecting
> > > >>> a new storage solution what are the important points that needs to
> be
> > > >>> looked into the vendors proposal ?
> > > >> Generally speaking a DASD box is not just a bunch of disks. It is
> DASD
> > > >> space + performance + embedded software.
> > > >>
> > > >> DASD space - obvious
> > > >>
> > > >> performance - a little bit more complex. There is no single number,
> > like
> > > >> car velocity. Even cache size should be understood, because vendors
> > have
> > > >> different way of use of it. Two approaches: compare "guts" (cache,
> > disk
> > > >> modules, etc.) or compare benchmarks.
> > > >>
> > > >> embedded software - remote copy, point in time copy, some "obvious"
> > > >> things like SuperPAV, etc. Unless you have specific requirements all
> > > >> vendors apply. However their offering differ and it could be
> > important.
> > > >> And features like Safeguarded copy.
> > > >>
> > > >> ...and zHyperlink. All vendors have FICON32 nowadays, and AFAIK all
> do
> > > >> have zHyperlink.
> > > >>
> > > >>
> > > >> HTH
> > > >>
> > > >> --
> > > >> Radoslaw Skorupka
> > > >> Lodz, Poland
> > > >>
> > > >>
> > >
> > > ----------------------------------------------------------------------
> > > For IBM-MAIN subscribe / signoff / archive access instructions,
> > > send email to [email protected] with the message: INFO IBM-MAIN
> > >
> >
> > ----------------------------------------------------------------------
> > For IBM-MAIN subscribe / signoff / archive access instructions,
> > send email to [email protected] with the message: INFO IBM-MAIN
> >
>
> ----------------------------------------------------------------------
> For IBM-MAIN subscribe / signoff / archive access instructions,
> send email to [email protected] with the message: INFO IBM-MAIN
>

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to