RE: physical / virtual tape backup file at a time

[Evan Harris <spanner@xxxxxxxxxx>]

Hi Adrienne

You will probably find the save process is quicker and more space 
efficient if you go the SAVRST* route using object connect, albeit 
that it might take you a little more time to set up and get working.

In your case, FTP requires that you save the file to a SAVF first, 
thereby using your valuable limited disk space, and then send the 
file. If you do go this route be sure to use DTACPR(*MEDIUM) as 
suggested by John as it can save quite a bit of space and will also 
reduce your transmission time.

On the other hand, the SAVRST* operation will use some disk space 
while being transmitted, but my observations are that it does not 
make a complete copy of the object prior to transmission but instead 
as each chunk is saved it is transmitted and the space re-used for 
saving the next chunk; essentially the space required for the 
operation is whatever the buffer size is that SAVRST* uses. You will 
probably also save some time as the transmission and save operations 
are occurring more or less in parallel - as one chunk is being saved, 
another is being sent.

Last but by no means least, it is much easier to monitor for errors 
in a CL using the SAVRST* commands - using FTP as you have to scan 
the transmission logt afterwards. You could use Scott Klement's 
FTPAPI to capture errors in line (and add retry logic if necessary) , 
but the work to do this will probably outweigh the effort required to 
set up any ANYNET link to support using SAVRST*.

When the data arrives on your system is there any kind of space 
limitation at your end ? Can you safely accommodate all the data locally ?

You should consider carefully what the expectation is for this backup 
data in terms of a recovery situation. I would be reluctant to 
recommend this as any kind of reliable backup approach, there are 
just way too many things that could go wrong. I don't know how many 
objects are involved, but if there are more than a couple of dozen, 
then you are going to have quite a backup window going on to save and 
transmit the objects; they will all have to be saved at a common time 
to be used to restore a whole library for instance. It goes without 
saying that while you are saving and transmitting the files the 
application will be down. If the save files on your system are meant 
to be the primary recovery data then I think you have quite an exposure.

That's not to say none of this can work, like anything it can be made 
to work  and probably work reasonably well with sufficient effort, 
but it will more than likely require constant monitoring. I like my 
backups to be as simple as possible - the simpler they are the more 
likely they are to work if I have to perform a recovery. if  however 
this is just a back-stop measure and you are not required to sign off 
on recovering the system then go for it.

Regards
Evan Harris

At 03:58 a.m. 14/03/2007, you wrote:

> Joe/Jerry,
>
> Are you suggesting we backup one library or groups of libraries at a
> time and push it to the DR system?
>
> If we're going to write our own backup system I can transfer one SAVF
> file using FTP without any additional requirements.
>
> Our desire is to back up to tape piece meal - file by file or lib by lib
> using only tape and virtual tape - but it is sounding like this will not
> be feasible.
>
> It is beginning to sound like we will have to go by Joe's suggestion of
> FTPing save files consisting of groups of files.  This was actually one
> of the backup scenarios we were considering, but we were hoping that
> vitual tape back up would have been more flexible.
>
> I thank you Joe and Jerry for your suggestions and welcome any
> additional comments.
>
> Have a great day!
>
> Adrienne McConnon