[Info-vax] Meditech in the news

[Arne Vajhøj]

On 1/14/2022 12:18 PM, Bill Gunshannon wrote:
> On 1/14/22 8:51 AM, Simon Clubley wrote:
>> On 2022-01-13, <kemain.nospam at gmail.com> <kemain.nospam at gmail.com> wrote:
>>> One biggie for tape is that current LTO tape solutions will do a 
>>> write, then
>>> read-back for every write. Disks just do a write, unless you do a
>>> backup/verify type operation.
>>
>> That only guarantees that the data, as presented to the tape drive, is
>> what was actually written to the tape.
>>
>> It does nothing for a failing controller delivering the wrong data to
>> the tape drive or for a faulty backup program delivering the wrong data
>> to the controller.
>>
>> IMHO, you still need a full verify pass by the backup program.
> 
> COOP!!
> 
> 40 years ago before any of this OOP or Agile crap took over
> the IT world I worked doing COBOL in a Univac 1100 environment,
> Once a month we used to pack up a backup set and travel to Rome
> AFB which was our designated COOP site.  We would take over one
> of their 1100's for the morning and do a complete restoration
> of the OS and all our data from the backup tape. Does anyone
> still do this?  Does anyone here even know what COOP means
> (without Googling it!!)

Restore test is still considered necessary and full
recovery test based on tapes and iron is still considered
desirable.

But a full recovery test is often difficult
today.

The Kyoto super computer that triggered this thread
is in an environment with 3 HPC clusters with
1800, 850 and 16 nodes respectively and a storage
system with 24 PB.

In a worst case where data does not compress well
then that is like 1000 LTO-9 tapes and 10000
tape station hours to restore.

I am pretty sure that they never did it.

Arne

PS: And no - I do not know that FLA.