|
Editor's intro:- Will
disk to disk backup make tape backup obsolete?
That's a
question that's been debated hotly here on StorageSearch.com for
many years. At the
extreme polarized ends of the argument are tape media makers like Sony, who in
an article here made a
case for the long term survival of tape, and at the other end of the argument
are disk to disk supporters like StorageSearch whose
editorial view has been
that tape doesn't have a viable role the midsize market any more. In the
middle of this argument are the moderates who say that maybe tape and D2d can
co-exist. This article by Steve Gardner at Engenio takes the middle course line
- and says why he thinks there's still a place for both. See if you agree. | |
(this classic article below was published
here at this url in December
2004)
|
|
Tape drives
increasingly take 2nd place when it comes to backups.
Inexpensive ATA
hard disk drives and
new appliances are stirring up considerable interest in disk-to-disk backup
technologies. However, only when both the hardware and software form a carefully
tuned system, do customers reap the benefits of enhanced
reliability and
speed when making their backups, according to Steve Gardner, Product Manager at
Engenio Information Technologies.
Although backup
has not risen to become the star task among IT departments in the last couple of
years, since September 11, 2001, its smooth functioning is, however, better
appreciated. According to IDC
in January 2004 the business issues surrounding data protection and disaster
recovery are rapidly becoming the most important market drivers behind the
growth in new storage capacity.
Despite the fact that many CIOs continue to regard the task of
creating backups as a nuisance and leave its running and implementation to the
corresponding specialist departments, the role of backups has transcended from
pure disaster prevention to become an integral part of the information
management process within companies. What is more, an increasing number of
providers of storage and backup solutions with ILM (Information Lifecycle
Management) are appearing on the market. If backup is to be regarded as an
integral part of the storage architecture, it must satisfy two requirements,
namely that of responding flexibly to the demands of administrators and users
and that of being able to be controlled centrally. It thus comes in very handy
if backups are no longer carried out on tape media like an irrefutable law, but
can integrate faster and more flexible storage media. Disk-to-disk backup is
the name given to the latest trend in the backup sector, which is being
propagated by analysts, manufacturers and users alike.
Disk as the first stage in the backup pyramid
At first glance, disk-to-disk
backup looks like a new edition of Hierarchical Storage Management (HSM).
HSM has a multi-layered design and moves the files in line with the age and
frequency of their use gradually onto ever-cheaper media. Complexity and costs
prevented HSM technology from really taking off at the time of its launch.
Today, however, the situation has changed. On the one hand, prices for
high-performance disk systems are far lower, while on the other hand, "high
performance" means a great deal more now than what performance meant back
then.
What therefore distinguishes these two? - throughput and fail safety,
which just a few years ago, were only to be found in the mainframe sector. In
this sector, disk-to-disk backup systems are nothing new. Since 1998 Storagetek
has sold its Virtual Storage Manager (VSM) almost two thousand times in
mainframe environments. However, the Open Systems sector is quickly catching up.
Companies do not regard disk-to-disk backup primarily as a means of
reducing costs, but rather as a flexible intermediate layer for backing up data
faster and, above all, for accelerating the restore process of current data.
Competitive comparisons between disk-to-disk backup and
conventional tape backup are constantly being made in the trade press and among
manufacturers - particularly among manufacturers of storage systems, but
customers do not see it like this. For users who don't want to rely solely on
hard disk drives disk-to-disk backup can fit into corporate networks as a
supplement as opposed to a replacement to tape. An indication of how far
disk-to-disk backup has been accepted is reflected by the fact that government
agencies are already eyeing up this new form of backup technology. For example
the US Securities Exchange Commission (SEC) published an update to its 17a-4
directive in late-June, in which disk-based systems were permitted to be
deployed as backups provided they met specific requirements.
Consortium pushes disk as backup medium
For the manufacturers of backup systems, the topic of disk-to-disk
backup has been around for long time. In the spring of 2003, a number of storage
manufacturers came together to form the Enhanced Backup Solutions Initiative,
which itself is now a member of
SNIA. Its members include
ATTO, Avamar, LEGATO Software, Network Appliance and Quantum. Besides providing
support for the members' products in a virtual tape environment, the objective
is to convey to users the potential of using disks as a backup medium. In doing
so, the companies are kicking at an open door in many places. From a technical
viewpoint, the advantages of using a disk drive as a backup destination are
plain to see.
Depending on the storage system, backup times are noticeably and even
dramatically shorter. In many companies, this technology renders the backup
window a lot smaller. The outcome being that backups are more up-to-date, since
they can be run more frequently, while application and backup servers become
available more quickly for other tasks and are under less strain. For many
companies, this alone is reason enough to contemplate extending their backup
system with a disk-to-disk backup solution. What is more, the advantage of a
random access storage system becomes all the more apparent when it comes to
restoring data. The idle times required for mounting tapes, searching for the
correct position, winding forwards and back are all reduced virtually to zero
under current LogicStor solutions. Tests by the backup software specialist
Veritas show disk-to-disk backup as being seven times faster, while data could
be restored even 20 times faster. Needless to say, the hardware used here plays
a significant role, and companies that demand short restore times generally use
high-performance backup and storage systems anyway.
Babylonian myriad of terms
Without a doubt, disk-to-disk backup is currently a hot topic,
which is attracting a great deal of attention. And, at the same time, the
interpretations are running wild. In a similar way to that witnessed recently
with the storage buzzword "Virtualization", when considering the topic
of "disk-to-disk backup" manufacturers and customers are often
thinking about different things. Roughly speaking, it is possible to categorize
the technology into three individual areas. On the one hand, there is the
disk-to-disk backup variant that every backup software masters anyway. The
programs have long been able to address not only tape media as the backup
destination, but also hard disks.
The advantage here being integration into the existing backup system.
Professional solutions such as Veritas and Tivoli treat a save set on disk no
differently to a save set on tape. Backing up, restoring, searching - everything
is performed within the storage management software application. This method
produces vast increases in performance, since multiple save sets can be written
onto disk simultaneously. In addition, it is possible to start the restore
operation as soon as a save set has been completed, even while other save sets
are being saved on disk. In this case, disk means both those hard drives
installed locally within the server or connected via
SCSI, and also network
drives. With one difference, however - whereas all file systems (generally
speaking, NTFS with Windows and ext3 with Linux servers) are supported on local
data media, backup programs are more discriminating in the case of network
drives, which often have to be NFS if they are to be supported as a backup
destination.
Once the data has been written onto this fast storage medium, there
are two methods in which to proceed. With the staging method, the backup
software writes the save set from the disk as soon as possible onto tape. Once
the software has completed this and has verified the backup, it deletes the save
set from the hard disk. With cloning, the data initially remains on the disk,
where it is available for rapid restore operations, although it does take up
space on the disk-to-disk backup system.
Both methods pose exacting requirements in terms of the reliability
of the disk system. Consequently, users are resorting to powerful and highly
redundant arrays, such as those based on technology from Engenio Information
Technologies, for example. Within its technology platform, Engenio offers
solutions designed specifically for back-up to disk based on fibre channel
arrays that are Serial ATA ready.
Meta Group regards this disk-to-disk backup variant as offering the
greatest potential for scalability and estimates that from 2004 to 2005 numerous
data protection suites will appear on the market that integrate disk-to-disk
backup even further. After all, thus far, not all backup programs have been
equally well suited for handling disk-to-disk backup.
However, there is a growing tendency to integrate automated routines
into the software. They manage the data-to-disk and data-to-tape transfer by
policy driven routines. This sounds familiar. HSM tried to achieve the same
effect almost ten years ago. The main difference lies within the seamless
integration of those routines into the backup programs, often dubbed "Next
generation" backup software. The most important advantage: it requires no
changes to the backup infrastructure, effortlessly drawing disk and tape
resources into the backup system without requiring special, dedicated hardware.
Other additional benefits of this model are the obvious protection in
existing investments and the fastest available random read/write times. On the
other hand, the second variant, dedicated backup appliances, also known as
Virtual Tape Libraries (VTL), rely on special hardware. These pass themselves
off to the backup program as a tape drive or a library. Quantum's DX30 and DX100
fall into this category. The greatest advantage being that implementation is in
many cases quick and easy, and the backup software does not even have to be
supported. The administrator configures as to which library, the number of slots
and which drives the VTL is to announce itself. For the backup application,
everything else is transparent, and does not "notice" any difference
between an LTO and a DLT drive. The majority of the appliances incorporate
redundancy and features for high availability such as RAID controllers, multiple
power supplies and specially adapted and stable operating systems.
Fibre channel
is used as an interface, which avoids bandwidth problems. In the same group, but
at the top end of the performance scale, you will also find such products as
CentricStor from Fujitsu Siemens and Storagetek's Virtual Storage Manager for
mainframe environments as well as its derivative for Open Systems, namely
EchoView. EchoView works with snapshots, complete images of a volume, which are
written onto the appliance and reserved until they have exceeded the set age or
the space on the disk runs out. CentricStor is certified both for Windows and
Solaris environments as well as for large mainframe systems. The solution acts
as a disk-based front end for the backup software and conceals the
tape libraries connected
behind this. Access by more than one host at the same time is thus also possible
as is the parallel backup of data to multiple backup destinations. LEGATO's
NetWorker supports CentricStor as a standard feature and maps the appliance like
an ordinary tape drive. The appliance is transparent for all the management
functions of the media, with the NetWorker users only seeing save sets,
irrespective of whether these are still in the cache of CentricStor or have
already been written onto a genuine tape drive.
Snapshot-based backups work best in combination with versions one or
two. In contrast to the conventional method, in which all the files to be backed
up are transferred to the target medium one after another, a snapshot creates a
virtual copy of the volume. Virtual because the data blocks remain untouched,
only the details of their position in the file allocation table (FAT) being
duplicated. The benefits are obvious for all to see: The process lasts from just
a few seconds to a few minutes, depending on size.
Application availability is impacted only for that, very short amount
of time. Compared to a traditional backup which can take hours, users have much
more access to applications and information. In addition no storage space is
required initially. Only when users access data blocks and change them, the
system retains a copy of the original blocks as produced at the time of the
snapshot. At this point additional storage space is required, amounting as a
rule to between 10 and 20 % of the overall volume per snapshot.
As the snapshot is completely decoupled from the actual data, it
ideally suited as a source for backups. Open files are no longer a problem, and
even active databases can be backed up in this way.
Snapshots can be implemented at three places in the storage network.
Firstly on the applications side, as a part of the host operating system or
application software. Another possibility is within the SAN, typically as a
router or other device located within the SAN fabric. In the case of the last
and most commonly used version, the storage system itself initiates the
snapshots. According to IDC, over three quarters of all snapshot implementations
are based on the storage system. In this context Engenio offers SANtricity
Snapshot. This software component of the storage system works fully
self-sufficiently, with the result that the host experiences no additional
loading. The virtual volume can be used as a backup source, for rapid backup
storage when changing versions and for testing new software without restricting
access to the actual volume of data. Even active databases can be backed up in
this way - the system merely has to suspend writing operations for the duration
of the snapshot.
Furthermore, when combined with LEGATO PowerSnap modules, this process
is fully automated from the backup software. Databases do not have to be
stopped manually, suspended during snapshot and restarted by hand again. LEGATO
SnapShot provides full snapshot 'lifecycle management' by a policy driven menu
for snapshot creation, expiration and management.
Disk-to-disk backup important for restore performance
The third method of implementing disk-to-disk backup systems
greatly simplifies the restore process. These systems comprise both hardware and
software and have the task of saving data as fast as possible from the source
system onto the appliance. The appliances achieve this with the aid of their own
software and compression procedure, but still retain the file format of the
original information. For a restore operation, the user can choose each
individual file or entire directories and write these back without a backup
software having to be involved. NexSAN's ATAboy or ATAbeast in conjunction with
InfiniSAN D2D software fall into this category. In principle, this entails a
staging procedure, as the data are to be written from the appliance onto tape
with the aid of a backup application. An incremental backup onto disk is
conceivable here, with a complete backup onto tape being performed just once a
week. Backup, and above all, restore procedures thus run very quickly, although
the software is still an additional unknown factor on the server system.
Traditional backup software such as LEGATO's NetWorker has been on the market
much longer and constitutes a more familiar factor for the users. It is also
necessary to keep an eye on the bandwidth of the connection from the backup
appliance to the host, since this is generally implemented via Ethernet.
Peaceful coexistence
Quite evidently, disk-to-disk backup is a complimentary technology.
Apart from a few exceptions, customers, manufacturers and analysts agree that
users cannot and indeed do not want to do without a further, external security
layer. There are several reasons for this; tapes have the edge in terms of
simple transport. Even if hard disk drives were designed to be mechanically
exchangeable, they are far more sensitive in terms of transport than tape media.
And naturally tape media still enjoy a clear price advantage per megabyte or
gigabyte - depending on the calculation model, the costs for a megabyte are less
than half a cent for tapes. Even very low-cost disk systems cost around 2.5
cents, with special Disk-to-Tape appliances greatly exceeding this, costing
between four to nine cents.
Nevertheless, these figures are a long way off the tape/disk cost
factor of recent years. Should this trend continue in the same direction,
Disk-to-Tape solutions could well be on a par with tape media in the medium
term, thereby equalizing the price factor. When it comes to making their
purchase decision, users should therefore always take account of the possible
savings due to significantly reduced downtime. Disk-to-disk backup systems are
faster at restoring data too. This means users become productive again
faster and the costs of a system failure diminish. ...Engenio profile
|
|
If you liked this article - take a look at
some other articles written by the same author
Here below are some
later related articles written by the editor (me)
what's the state of DWPD? Reaching for the petabyte
SSD where
are we heading with memory intensive systems? How the backup market moved
from tape to disk and VTLs |
|
|

| |
|
. |
Old storage news Reaching for the petabyte
SSD Cloud
adapted memory systems Classic Storage
Architecture - guide Where are we now
with SSD software? How fast can your SSD
run backwards? How
the backup market moved from tape to disk Segmentation
problems and user needs in enterprise SSDs |
. |
 |
. |
"When you're looking
at storage on the SAN then by the time we get to 2016 (plus or minus a bit)
then you won't need a petabyte of SSD to replace a petabyte of HDD. The
replacement ratio will be somewhere between 10 to 1 and 50 to 1..." |
the enterprise SSD
software event horizon | | |
. |
|
. |
the Problem with
Write IOPS
the "play it again Sam" syndrome |
Editor:- Flash SSD "random
write IOPS" are now similar to "read IOPS" in many of the
fastest SSDs.
So
why are they such a poor predictor of application performance?
And
why are users still buying
RAM SSDs which cost
9x more than SLC? - even when the IOPS specs look similar. |
 |
This article tells you
why the specs got faster - but the applications didn't. And why competing SSDs
with apparently identical benchmark results can perform completely
differently. ...read
the article | | | |
... |
Fast Purge flash SSDs -
when "Rugged SSDs"
won't do |
The need for fast and
secure data erase - in which vital parts of a flash SSD or its data are
destroyed in seconds - has always been a requirement in military projects.
|
 |
Although many industrial
SSD vendors offer products with extended "rugged" operating
environment capabilities - and even
notebooks SSDs come
with encryption - it's the availability of fast data purge which
differentiates "truly secure" SSDs which can be deployed in
sensitive applications.
| | | |
... |
|
... |
|
. |
after AFAs -
what's the next box? |
Editor:- In the past we've
always expected the data capacity of memory systems (mainly DRAM) to be much
smaller than the capacity of all the other attached storage in the same data
processing environment. That will change. The growing use of new memory centric
apps like analytics and machine learning means that legacy storage systems will
be a small part of the memory systems capacity installed locally - even when all
the storage is SSD based. Memory is the new storage. | | |
. |
SSDs - the big picture |
Editor:-
StorageSearch.com was the
world's 1st publication to provide continuous editorial coverage and analysis
of SSDs (in 1998) and in the 12 years which have followed we've led the market
through many interesting and confusing times. |
 |
If you often find yourself
explaining to your VC,
lawyer or non technical BBQ guests why you spend so much time immersed in SSD
web pages - and need a single, simple, not very technical reference to
suggest - this may be the link they need. | | | |
. |
 |
. |
SSD Pricing -
where does all the money go? |
SSDs are among the most
expensive computer hardware products you will ever buy.
Understanding
the factors which determine SSD costs is often a confusing and irritating
process... |
 |
...not made any easier when
market prices for identical capacity SSDs can vary more than 100x to 1!
Why is that? ...read
the article | | | |
. |
SSD encryption - does
it hurt? |
Editor:- a reader asked me a
good question...
Does SSD encryption hurt performance and
reliability? |
| | |
. |
|
. |
 | | |