Need Suggestions for Automated Backup

[Chong Yu Meng]

I know of a video-editing company in Brunei that uses RAID-5. I'm not 
too concerned about speed because my feeling is that the network will be 
the bottleneck. In any case, RAID-5 will require me to purchase 3 hard 
disks and I'd still need a backup system. I think I'll go for the 
mirrored hard disks first, maybe move to RAID-5 when I have the budget, 
but I'd get a backup tape drive first before moving to RAID-5.

I had some really hairy experiences with tape backups years ago -- can't 
restore backups, can't successfully backup database files, etc. But very 
recently, when I was using a Sun backup drive (I think it's actually a 
Quantum), it actually worked very well. Backup speeds were pretty 
impressive (1 GB in 15 minutes), and the backups actually worked -- i.e. 
they could be restored. So, I'm ok with tape.

I guess my problem can be solved procedurally : start small, slowly 
build up capability as volume ramps up (and the tall dollars roll in -- 
hopefully!).

I miss the dot-com days when I had a huge budget and could build a whole 
server farm !

Thanks to All !

pascal chong



Alma J Wetzker wrote:

>
>
> Matthew Carpenter wrote:
>
>> From: "Net Llama!" <netllama at linux-sxs.org>
>>
>>> Its worth noting that RAID5 is generally alot slower than RAID0 or 
>>> RAID1.
>>
>>
>>
>> That's not been my experience.  Using hardware RAID at least, RAID1 
>> is much
>> slower than RAID5.  RAID0 doesn't count since it is only used for 
>> speed and
>> doesn't provide data-protection.  In that way, RAID0 is a misnomer.
>> RAID1 is a mirror, so the data is being written and read much like a
>> standalone drive, assuming you have a separate controller for each 
>> drive.
>> If both drives are on the same controller RAID1 is even worse.  RAID5
>> spreads the data out so reads are (can be) much quicker than a RAID1 as
>> multiple tracks can be read at virtually the same time.  Writing is 
>> similar.
>>
>> I'm not sure how SW RAID5 compares to SW RAID1.
>
>
> Please correct me if I am wrong, I have been out of this stuff for a 
> few years.
>
> I don't know of anyone using just RAID1 or just RAID0.  The 
> installations that I know/knew about used RAID (10) or mirrored 
> stripes.  That puts lots of spindles over your data for very fast 
> reads and writes.
>
> RAID 5 is strips with checksum.  That allows a single drive in the 
> array to fail and you still get all your data.  If you have hot 
> pluggable drives, you can replace a failed drive on the fly.  The 
> reads are not quite as fast as RAID (10) because you only have half 
> the spindles.  The writes are much slower because you need to generate 
> the checksum for the file.  If you are updating a database, the files 
> can be large and the checksum calculation long also.
>
> If you do this with software rather than hardware, your OS will incur 
> a huge performance hit with RAID 5 because of the checksum issue.  
> (You need to read the entire file for every update.)  Also remember 
> that IDE drives have the basic inteligence of a floppy drive.  If you 
> want optimized retrieval at the hardware level, get SCSI.
>
> My experience is that doing this stuff in hardware is best, (Duh!) and 
> that RAID 5 reads data about as fast as RAID (10).  The selection 
> criteria then becomes what is more important, write speed (with cost) 
> or just uptime of the array (cheaper).  NOTE: Both are faster on reads 
> than a single drive alone.
>
> I know everyone knows this but this time I knew it too ;)
>
>     -- Alma
>
> _______________________________________________
> Linux-users mailing list
> Linux-users at smtp.linux-sxs.org
> Unsubscribe/Suspend/Etc -> 
> http://smtp.linux-sxs.org/mailman/listinfo/linux-users
>
>