Original question :
> > We're planning to connect a couple of AlphaServers (running DU 4.0B)
> > to a EMC Symmetrix 3430, for performance and compatibility testing.
> I found no summaries on this topic, so I'd like to know if you folks
have
> some experience with this beast. Hint and gotchas are most welcome.
Many thanks to :
Dr. Tom Blinn
Stefan Moeding
Mike Sullivan
Jon Morgan
Alexander Sakellaris
Viktor Holmberg
They were very helpful and probably I will bug them
again in the future ;-)
The main points to check :
1) The Symmetrix array can handle 15 devices on the FW SCSI chain,
but Digital Unix only supports 7.
2) A Symmetrix entry on the /etc/ddr.dbase may be needed.
3) The disktab entry may not be needed since the dynamic geometry
labels can be used.
4) There are some special considerations for ASE (we're not using
it anyway)
5) It's a must to have a microcode release which supports stripped
metavolumes (we will have it).
6) EMC supports dual paths to the disks, so you could access the
same device through two SCSI channels. DU doesn't understand this
(not until 4.0F according to Viktor), so one may find different
devices
referencing the same physical unit, which is potentially a bad
thing.
The solution is to ask EMC to remove the secondary paths.
Thanks again,
Miguel Fliguer - MINIPHONE S.A.
Buenos Aires, Argentina
> ----------
> De: Holmberg, Viktor[SMTP:Viktor.Holmberg_at_abnamro.co.uk]
> Enviado el: jueves 31 de diciembre de 1998 9:44
> Para: 'Fliguer, Miguel'
> Asunto: RE: Symmetrix on DU 4.0B, anyone ?
>
> Excellent equipment. You can get StorwareWorks to go faster with
> HSZ70,
> but you don't get the extra features, SRDF , TimeFinder etc.
>
> OK, trick and tips.
>
> To mirror or not to mirror that is the question. Do you want to mirror
> with LSM or via the Symmetrix ?
> - Symmetrix will automatically replace a failed drive, no intervention
> from system manager, like hotspare but will do it predictively, it
> monitors the errors from the disks and swaps it out BEFORE it has
> failed.
> - Symmetrix will buffer all mirrored writes reducing the write time.
> - Symmetrix mirroring reduces overhead on CPU.
> - Symmetrix mirroring allows you to reduce the write bandwidth
> required
> on you scsi buses, increasing your available bandwidth (double).
> - Symmetrix mirroring places the scsi bus as a single point of
> failure.
> I have never had a scsi card fail on me yet (touch wood).
>
> * I recommend to mirror in Symmetrix and stripe across as many scsi
> buses you can.
>
> As you will be aware Symmetrix use 9/18/21 Gbyte disks. They normally
> slice these disks in to 4 Gbyte hyper-volumes which is reasonable.
> This
> is because the Symmetrix will create a disk queue for each
> hyper-volume
> and hence can perform multiple write to a disk. These hypervolumes
> have
> to be managed properly. Make sure that you split your mirror sets and
> stripe sets across physical disks. EMC will try to dictate to you how
> you should do things or they will do it automatically. They don't know
> that you want to do with the volumes later. If you are connecting
> multiple machines to the Symmetrix make sure they don't give you
> hyper-volumes that clash with other machines. At one site EMC
> configured
> two separate application on the same physical volumes (via hyper
> volumes), one application had 2/3 of the disk the other 1/2. The
> smaller
> application died. The batch time trebled.
>
> Make sure when then create the hyper-volumes that they tell you what
> internal scsi bus (DA disk adapters or back-end) and disk are
> connected
> to. You should tell them what external host bus (CA channel adapters
> or
> front-end) they should be seen on and what scsi-id and lun they should
> have. Each Channel adapter has 4 scsi buses and is connected to one of
> the two internal backplan. So for maximum performance make sure you
> stripe across both CAs, or if you are mirroring mirror across the
> different CAs for resilience and performance.
>
> EMC normally work with HP systems who have dual paths to the disks,
> i.e.
> you can see the same disk down different scsi channels and if one of
> the
> patch fail LVM automatically uses the secondary path. This feature is
> available in Digital UNIX 4.0F. As a result they, EMC, will normally
> create the secondary patch for all the disks. However Digital UNIX
> don't
> understand dual paths and sees the disk as two separate drives. If you
> access the secondary path of a disk that you are already using it will
> corrupt your disk. Request that EMC remove all secondary paths. It
> will
> save problems in the future.
>
> If you have any questions let me known and I will try to help.
>
> Cheers
> Viktor
>
>
>
>
> > ----------
> > From: Fliguer, Miguel[SMTP:M_Fliguer_at_miniphone.com.ar]
> > Sent: 30 December 1998 18:20
> > To: 'osf-list'
> > Subject: Symmetrix on DU 4.0B, anyone ?
> >
> > Regards,
> >
> > Miguel Fliguer - MINIPHONE S.A.
> > Buenos Aires, Argentina
> >
> > PS. I hope everything is fixed on our mail server and you get this
> > message without the annoying HTML trailer. If itīs still there, just
> > let me know. I apologize in advance.
> >
> >
> >
> >
>
>
>
> -------------------------------------------------------------
>
> This email and any files transmitted with it are confidential
> and intended solely for the use of the individual or entity
> to whom they are addressed. This communication may contain
> material protected by attorney-client privilege. If you are
> not the intended recipient or the person responsible for
> delivering the email to the intended recipient, be advised
> that you have received this email in error and that any use,
> dissemination, forwarding, printing, or copying of this email
> is strictly prohibited.
>
> If you have received this email in error please notify the
> IT department at postmaster_at_abnamro.co.uk or by telephone
> on +44 (0)171 601 0101.
>
>
>
Received on Tue Jan 05 1999 - 15:18:01 NZDT