Distinct backup-to-disk architectures emerge
Salvatore Salamone, FRAMINGHAM
As database volumes grow larger, backing up data
to disk, rather than tape, is gaining favour among
life science companies. Now, as the backup-to-disk
field develops, companies are finding that there are
a couple of approaches that may be taken, each with
different costs, benefits, and trade-offs.
Discussion of the various backup-to-disk approaches
was the focus of a recent talk by W Curtis Preston,
president and CEO of the consulting firm The Storage
Group. Preston's talk, titled 'Integrating Disk into
Backup for Faster Restores,' was delivered at the
Storage Decisions conference held in New York last
week.
"Tape backups are not cutting it," says
Preston. He notes that many are finding it simply
takes too long to perform backups and to restore files
when dealing with larger and larger databases.
For example, Beyond Genomics found that as its database
grew, tape backups were exceeding their 48 hour, weekend
window allocated to perform such tasks.
Additionally, companies are often not following good
business continuance/disaster recovery practices of
making a copy of each tape to store off-site. "Tape-to-tape
copying takes (a long) time, so companies are not
copying original backups and sending them off site,"
says Preston.
This point was reinforced by an impromptu poll by
Preston. Using a wireless audience response system
at the conference, 7 % of the attendees said they
are not sending anything off-site and 46 % are simply
shipping the original tapes off-site without making
copies. Less than half (49 percent) of the attendees
said they make copies of original backups and send
a copy off-site.
The traditional solution to deal with both the increasing
time to perform a backup and the need to make copies
of tapes for safe off-site storage has been to add
tape drives. By adding tape systems, backup can be
spread over multiple tape drives increasing the aggregate
throughput and thus, shortening the time required
to perform a backup operation. Such additional drives
can also be used to make copies of original backup
tapes.
This solution is not cheap since it means buying
more tape systems and results in many tape drives
sitting idle when backups are not being performed.
In fact, many companies have increased the capacity
of their tape systems to the point where they simply
are not efficiently using them.
Enter backup-to-disk
For these reasons, backup-to-disk is getting closer
scrutiny in many life science companies today.
As IT managers look at the backup-to-disk alternatives,
two distinct architectures are emerging, according
to Preston. Both have similar physical configurations
that involve backing up essential data stored on high-performance
disk-based systems to a secondary disk-based system,
which then connects to tape drives.
In one scenario, the secondary storage system essentially
acts as a cache that holds the backed up data until
it is written to a tape. In this scenario, most restores
are done from tapes. And the data on the secondary
storage system might only reside on disk until it
is backed up tape.
This approach to backup has a couple of benefits.
First, the secondary disk-based storage system does
not have to be enormous since the system only holds
data until it is put on tape.
Second, in contrast to backing up data directly to
tape, this approach buffers the data before it hits
the tape drives. So rather than streaming data to
multiple tape drives (which is the common technique
used today) to get a higher aggregate throughout to
meet a backup time window, fewer tape drives might
be used.
In the other emerging scenario, all data is backed
up to the secondary disk-based system, which is also
used to restore any lost data. Tapes are used to backup
this secondary storage system and are only used in
disaster recovery situations.
One benefit to this approach is that all backed up
data is online. Restoring files or data volumes is
typically faster with this approach since there is
no time lost looking for and fetching the appropriate
tape with the copy of the lost data. And restoring
a specific file is faster since the data does not
have to be read sequentially (as is the case with
tapes).
This second scenario is in great favour today. At
the conference, 75 % of the audience preferred this
method of backup. But Preston points out ,the only
issue here is cost." Even when using lower performance
disk systems for backup, the cost per gigabyte can
be about five times that of tape. However, Preston
says that he is starting to see backup-to-disk systems
coming to market with lower dollar per GB prices.
One final point from the talk is that tape is not
going away. While backup-to-disk offers higher performance
and faster times for straight backups and restores,
Preston notes that tape will still be used in conjunction
with these systems to provide archiving of essential
data and to support business continuance/disaster
recovery initiatives.
