Disk geometry in OpenVMS

[Ronny_7]

Hi,

I have two different set of OpenVMS cluster running on OpenVMS V7.1-2.

First cluster is accessing a HSJ50 storage while the second cluster is accessing a HSZ80 storage.

In these two storages, each has a stripeset consists of four 9.1GB disks.

In the first cluster connected to HSJ50, the show device full shows the stripeset has,
Total cylinders 5258
Tracks per cylinder 80

From the second cluster connected to HSZ80,
Total cylinders 21029
Tracks per cylinder 20

May I know what cause this difference and who determines these geometry?

Thanks and Regards,
Ronny

[Willem Grooters]

Ronny,

I have no experience with this kind of controllers, but can think of a few possibilities.

Assuming that the stripe-sets are created from the controller, I'm quite certain that the difference is caused by the controller. For reasons, I can think of optimalisation of seek, read and write operations, speed of the on-board porcessor, memory-size and organization - buffers and cache.
The actual geometry of the physical disk won't matter (but could have some influence), it's more the way the set is organized over these physical disks. That too _could_ be different between the two controllers! (making is impossible to change controller only, I think).

One other possibility is that internally the structures are equal but the result of teh request by VMS to send this geometrical information is interpreted differently by each controller.

[Uwe Zessin]

Ronny,

as Willem has already said, it is the controller who 'decides' what geometry to present. The HSJ and HSZ series have a somewhat different history and talk different protocols (MSCP / SCSI) to their hosts. One would have to ask the engineers of both controllers for the reason.

As far as I know VMS has given up most if not all dependency on disk geometries. Since VMS V6.2 the geometry-independend placement of homeblocks is possible (its the default). It doesn't even make sense any longer, because modern drive don't have a fixed geometry (cyl,head,sector) anyway and a RAID controller between the host and the physical disks will completely invalidate any assumptions.

[Ronny_7]

Hi Willem and Uwe,

Thank you for the detailed explanations.

Regards,
Ronny