Even though the sound of it is something quite atrocious, the
biggest storage issue isn't San, Nas, fibre or switches, it's cost.
This is where consolidation comes in, reports Nick Enticknap.
Consolidating storage makes obvious sense. You gain economies of
scale in hardware purchase; you make more efficient use of the
storage you run; and storage management becomes both easier and
cheaper. It's a no-brainer.
Gartner Group analyst Josh Krischer summarises the argument:
'Consolidating disk gives you better use of spare capacity. It also
reduces the cost of storage management. In future, this issue will
become bigger, because of the people costs and the explosion in
data capacity.'
A simple statistic from analysis company Sosinsky Group underlines
the point. The disk utilisation rate with server-attached storage
is 50 per cent; with consolidated storage, it goes up to 85 per
cent. Even with today's plummeting disk prices, the straightforward
hardware savings are substantial.
So why hasn't everybody done it? Because consolidating storage
involves a lot of careful planning and significant up-front
investment. Because there are various different ways of
consolidating storage, but all with disadvantages as well as
benefits. Because consolidating storage involves different thinking
about issues such as backup and restore and storage management
software. Because there are concerns about protection of existing
investment, and about getting locked in to any chosen new
solution.
Bloor Research senior analyst Tony Lock said, this means storage
consolidation 'takes time and effort, which are the two things in
shortest supply'.
Nonetheless, the arguments for consolidating storage are so
powerful that users are steadily moving in that direction.
According to Gartner Dataquest, traditional direct-attach storage
(Das) accounted for $24bn revenues out of a total disk storage
market of $32.4bn in 2000 - nearly three quarters. The analysis
firm predicts that Das sales will decline at 6 per cent a year over
the next three years, while all other types of storage product and
service will rise. As a result, Das will account for less than half
of 2003 disk sales - $20bn out of an $54bn market.
These figures show that storage consolidation is still in its
infancy, but gaining momentum fast.
There are fundamentally three different ways of consolidating
storage. The first to arrive was platform-sharing - sharing an
enterprise disk subsystem between servers running different
operating systems. This approach was pioneered by EMC, which
introduced the ability to share its Symmetrix mainframe systems
with open systems - first Unix, then Windows NT - in 1995. This is
resource sharing at the hardware level - the different servers are
allocated their own storage partitions within the box.
As time has gone by, new facilities have been added to such
systems, including hot swapping of failed components, upgrading
without bringing the system down, and the ability to upgrade on
demand.
The storage area network (San), which arrived towards the end of
the last century, extends this principle by hiving off the storage
from the servers altogether, and placing it in a separate network.
This reduces the processing load on the servers, and makes hot
swapping and dynamic upgrading easier. It also opens the way for
server-free and Lan-free backup.
The San, like the multi-platform standalone system, provides
resource sharing. In both cases storage needs to be allocated to
specific servers and files remain in their proprietary formats,
accessible only by systems equipped with the software to read
them.
Network-attached storage (Nas) differs in being designed for file
sharing rather than disk sharing. The idea is that PCs and
workstations can share files irrespective of the system they were
created on.
These three types of consolidated storage all have disadvantages.
Multiplatform devices are restricted in the number of servers they
can attach to, and also have relatively limited scalability. Sans
require a lot of infrastructure investment in cabling and switches;
also, the standards-making process is far from complete, which
makes putting the pieces together a complex process. Nas systems
are designed to be connected to existing Lans, and can therefore
impose a strain on the Lan.
Much of the activity in the storage market today is aimed at
overcoming these disadvantages. There is a movement to converge San
and Nas. There are developments to build Sans from existing local
area network technologies such as Ethernet, rather than requiring
special purpose Fibre Channel components.
Mixed storage goal
According to Brocade's European marketing manager Paul Trowbridge:
'Today people support mixed servers but not mixed storage.' One way
of moving towards that goal is virtualisation, which makes it
easier to connect heterogeneous storage devices to a consolidated
system, and paves the way for file and database sharing.
There are many different ways of virtualising storage. It can be
done within the storage device, within the server, or within a
storage network. Suppliers in each category claim theirs is the
only sensible way of doing it, but that is nonsense. There are
advantages and disadvantages in all the approaches.
All this makes it difficult for the user to choose. For many, there
will not be a simple answer; two or all three of the options may be
applicable in different circumstances.
The prime need is for the user not to be distracted by the hype or
by the arguments and counter-arguments in favour of one approach or
another. They should concentrate on their own business objectives,
and select an expert partner to decide what technology will best
meet the requirements now and into the future. The immaturity of
the technology and the absence of many key standards makes using
external expertise necessary. It's not yet time to go it alone as
you can do with Lans.
Customer objectives
What should you do before approaching a business partner? Legato
system engineer Andrzej Dobrzynski comments: 'When designing a
solution you need first to establish the requirement for
information availability. You define customer service objectives -
system uptime; backup window; what data loss is acceptable; and
recovery time.'
This is why most market researchers are predicting a boom in the
market for this type of software. Gartner Dataquest is forecasting
growth from $5.3bn last year to $16.7bn by 2005.
One key aspect of storage management software is backup/restore.
This becomes more important as IT applications become more
mission-critical. Most attention is paid to backup, which is
reasonable as it is an everyday task. But it is becoming
increasingly recognised that the restore part also needs attention.
According to Gartner's Josh Krischer: 'We estimate that around 25
per cent of backups are useless because customers cannot
restore.'
It is worth noting too that backing up today's increasingly large
volumes produces a strain on the existing magnetic tape technology.
Even with smaller volumes, tape drives currently used for an hour a
day won't cope with being asked to run all day long in the new
environment. This is why tape drive/library suppliers such as
Overland Data and Quantum ATL are making their products more robust
and introducing fault-tolerant capabilities (Overland Data
guarantees 99.999 per cent uptime on its new LibraryXpress Neo
series launched in January).
Case study - Littlewoods
The Littlewoods Organisation decided to integrate its three trading
divisions - Home Shopping, Index Catalogues, and Retail - for
business reasons. The company decided to integrate the IT
infrastructure at the same time, to reduce maintenance and staffing
costs and to improve flexibility.
Littlewoods has adopted a networked approach, consolidating onto
EMC Symmetrix disk subsystems in two separate storage farms on
different sites, linked by SRDF (Symmetrix Remote Data Facility)
software. There are also two separate networks, built from Brocade
switches. The facility provides a total of 98Tb storage for over 50
servers, including an IBM MVS mainframe host and Sun Solaris and
Windows NT systems.
This infrastructure is now in operation, and has already produced
substantial benefits for Littlewoods. According to Group IT
Director David Hallett: 'I would estimate that in terms of
performance gain, saving in backup time, the pooling of Dasd and
reduced maintenance, Littlewoods is saving approximately £750,000 a
year.' He added: 'From the customer perspective the new solution
improves the uptime of applications and moves the company far
closer towards the goal of 24x7 operation.'