[DM-MUG] Backups (in response to the Danger/Sidekick Fiasco)

[Jon Thompson]

Victoria alluded to this in another post, but it requires some further  
explanation...

The fiasco with the Sidekick phones is apparently a result of poor  
planning on the part of Microsoft, and their SAN integrator- Hitachi.
(As a side note, I am a SAN admin, so I deal with these technologies  
on a day to day basis.)

Apparently Hitachi was updating the Storage Area Network (SAN,) which  
went south. Microsoft then went to their backup of the data, which was  
also corrupt. Updating a SAN is a relatively risky procedure, often  
requiring zero corruption on many tens of Terabytes of data. Because  
of this nature, it is imperative that a good backup is performed, as a  
heavily corrupt SAN is usually unrecoverable (a single file can be  
spread between not only many hard drives, but many devices.)

This provides the opportunity to talk about backups a bit...

1) A good backup is a verifiable backup

If you are dumping info into Time Machine, and you are _never_ pulling  
info back out of it and checking, how do you know that the data is  
good? In the instance of the SideKick data, this means that they would  
have known that the backup is good, and that they could rely on it.  
Until you know that you can recover from it, you cannot call it a good  
backup.

2) A good backup consists of many backups of the same data.

I store three months of backups, and I consider that too little, but  
don't dare increase my budget. If you are relying on all of your data  
being on a single hard drive, what happens if there is a fire and both  
your computer and backup is lost? If all of your backup is in one  
cloud (Mozy) and that cloud bursts like the SideKick cloud did, you  
are hosed. Now, if you have daily backups on a time machine volume,  
and a weekly backup to Mozy, you are protected from either example.  
Mozy is your offsite backup, the time machine

3) Synchronization is not backup.

Synchronization ensures that all of you data is the same between all  
instances that you create. If you are syncing your address book on  
your mac with an iPhone, you have two instances of the data, but they  
are tied together. If you delete a contact on one (or a virus deletes  
all of them) that deletion is transferred to your backup. I realize  
that there are ways that the sync service tries to reduce the chance  
of this, but it is still a possibility. However, an offline copy of  
your address book on a USB thumb drive that is not plugged into your  
computer cannot not be changed, at least until it is plugged in again.  
Dropbox is a synchronization service.

Part of the Sidekick debacle is because it is apparently a  
synchronization service. They are telling people not to power down  
their phones, because it apparently wipes the internal memory and re- 
downloads content from Microsoft servers. Of course there isn't any  
data to re-download...

4) RAID is not backup.

RAID (except 0) is a way to protect from hard drive crashes. It does  
not protect from application corruption, accidental deletion, or  
intentional deletion. All of which are much more likely than hard  
drive corruption during the majority of the life of a hard drive. RAID  
is useful for server admins that deal with distributing data amongst  
many hard drives, which actually is a great way to exponentially
increase the likelihood of data loss due to hard drive failure.

RAID 0 offers no protection at all- in fact, it is will exponentially  
increase your likelihood of data loss due to hard drive failure. Be  
careful when purchasing large external hard drive, as they are  
sometimes two drives that are internally configured as RAID 0. This is  
useful for professional video primarily.

5) Backup as sparse as you are willing to recover from.

- If you couldn't stand losing more than a week's worth of data, you'd  
best be backing up weekly.
- If you couldn't stand losing more than a day's worth of data, you'd  
best be backing up daily.
- If you couldn't stand losing more than an hour's worth of data,  
you'd best be backing up hourly.

You know what is important and what is not. Make sure to taylor your  
backup routine to your needs. Incremental backups, such as Time  
Machine are good about this.

6) Recovery options are available, but expensive.

Data recovery costs a minimum of $1000 for a hard drive. And they  
don't guarantee they can recover anything.

-- 
- Jon
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://cialug.org/pipermail/dmmug/attachments/20091013/ae0f1fc3/attachment.htm