The evolution of storage systems

of
hardware and software to obtain a highly
reliable, high-performance, and easily
managed system. We explain in this paper
how storage systems have evolved over ve
decades to meet changing customer needs.
First, we briey trace the development of the
control unit, RAID (redundant array of
independent disks) technologies, copy
services, and basic storage management
technologies. Then, we describe how the
emergence of low-cost local area data
networking has allowed the development of
network-attached storage (NAS) and storage
area network (SAN) technologies, and we
explain how block virtualization and SAN le
systems are necessary to fully reap the
benets of these technologies. We also
discuss how the recent trend in storage
systems toward managing complexity, ease-
of-use, and lowering the total cost of
ownership has led to the development of
autonomic storage. We conclude with our
assessment of the current state-of-the-art by
presenting a set of challenges driving
research and development efforts in storage
systems.
The rst data storage device was introduced by
IBM
in 1956. Since then there has been remarkable pro-
gress in hard disk drive (
HDD
) technology, and this
has provided the fertile ground on which the entire
industry of storage systems has been built. Storage
systems are built by taking the raw storage capability
of a storage device such as the
HDD
and by adding
layers of hardware and software in order to obtain
a system that is highly reliable, has high performance,
and is easily manageable. Storage systems are some-
times referred to as storage subsystems or storage de-
vices (although device is better used to describe the
raw storage component or an elementary storage sys-
tem). Originally, the storage system was just the
HDD
,
but over time storage systems have developed to in-
clude advanced technologies that add considerable
value to the
HDD
. Storage systems have evolved to
support a variety of added services, as well as con-
nectivity and interface alternatives. It is for this rea-
son that le systems and storage management sys-
tems are often considered parts of a storage system
and thus will be briey treated in this paper.
To understand the evolution of storage systems, it
is important to observe the evolution of the
HDD
.
The areal density of the
HDD
has improved by seven
orders of magnitude, and this has resulted in a re-
duction of the oor space taken by the correspond-
ing storage systems also by about seven orders of
magnitude. Figure 1 plots the
HDD
areal density (on
the left) and the price of various storage devices (on
the right) since 1980.
HDD
prices
have decreased by
about ve orders of magnitude since 1980, while the
cost of storage systems has fallen about 2.5 orders
Copyright 2003 by International Business Machines Corpora-
tion. Copying in printed form for private use is permitted with-
out payment of royalty provided that (1) each reproduction is done
without alteration and (2) the Journal reference and IBM copy-
right notice are included on the rst page. The title and abstract,
but no other portions, of this paper may be copied or distributed
royalty free without further permission by computer-based and
other information-service systems. Permission to republish any
other portion of this paper must be obtained from the Editor.
IBM SYSTEMS JOURNAL, VOL 42, NO 2, 2003
0018-8670/03/$5.00 2003 IBM
MORRIS AND TRUSKOWSKI
205 of magnitude in the same period.
1
The sharper fall
in the price of the
HDD
implies that the cost of the
raw
HDD
accounts for a progressively smaller frac-
tion of the total cost of the storage systemwe will
return to this crucial observation later.
Although Moores law tells us that the number of
transistors per unit area of silicon doubles every 1.5
years, we see from Figure 1 that the number of bits
stored per unit of
HDD
media is doubling about ev-
ery year! But more important than improvements
in device density or cost have been the new appli-
cations that have been enabled by these advances.
On the time line in Figure 1, two milestones stand
out. In 1996, digital storage became more cost-
effective for storing data than paper, and, in 1998,
we reached the point where lm used in medical ra-
diology could be economically supplanted by elec-
tronic means. Another important milestone, this one
in the consumer market, was reached several years
ago when it became cost effective to store video con-
tent using digital storage systems. Soon after we saw
the emergence of
HDD
-based set-top-box devices
(sometimes called personal video recorders) that of-
fered improved management of entertainment video
in the home.
Besides the emergence of storage systems that en-
able digitization and replacement of legacy media,
it is instructive to consider how architectural out-
comes are affected by the relative progress of tech-
nologies. Figure 2 portrays the relative advances in
storage, processor, and communications technolo-
gies, obtained by plotting the cost/performance im-
provement
of widely available technologies in end
user products as documented by personal computer
(
PC
) magazines since 1983. (Broadband to the home
is considered to have limited availability at present.)
The plot shows that since 1990, storage technology
has outrun both communications and processor tech-
nologies. In fact the availability of inexpensive dig-
ital storage has inuenced the architectures that we
see in place today. For example, because storage
technology has become relatively inexpensive while
deployment of point-to-point broadband to homes
has been slow,
HDD
-based set top boxes are more
prevalent than video-on-demand. The parallel story
of technology in the enterprise is not shown but leads
to a similar conclusion: the amount of stored data
has outrun the ability of communications and pro-
cessing systems to provide easy access to data. Thus,
we see widespread use of multiple copies of data
(e.g., Lotus Notes* replication, widespread internet
Figure 1 HDD storage density is improving at 100 percent per year (currently over 100 Gbit/in2). The price of storage is

decreasing rapidly and is now significantly cheaper than paper or film.
SINCE 1997 RAW STORAGE
PRICES HAVE BEEN DECLINING
AT 50% 60% PER YEAR
25%
CGR
100%
CGR
1980
1990
10 000
1000
100
10
1
0.1
0.01
0.001
1000
100
10
1
0.1
0.01
0.001
0.0001
HDD STORAGE DENSITY
STORAGE DEVICE PRICES
LAB
DEMOS
PRODUCTS
60%
CGR
PROGRESS FROM
BREAKTHROUGHS, INCLUDING
MR, GMR HEADS, AFC MEDIA
HDD AREAL DENSITY (Gb/in2)
PRICE/MBYTE (DOLLARS)
DRAM
FLASH
3.5" HDD
2.5" HDD
1" MICRODRIVE
RANGE OF
PAPER/FILM
1990
2000
2010
1980
1985
1995
2000
2005
2010
HDD
DRAM
FLASH
PAPER/FILM
MORRIS AND TRUSKOWSKI
IBM SYSTEMS JOURNAL, VOL 42, NO 2, 2003
206 caching) as well as the deployment of storage sys-
tems close to the end user in order to avoid network
delays.
We have already pointed out that the cost of the
HDD
accounts for a progressively smaller component of
the total cost of a storage system. In fact, in recent
years the game has changed not once, but twice. First,
the
HDD
has changed from a differentiating technol-
ogy to a commoditized component in the typical stor-
age system. As shown in Figure 6 of Reference 1,
the
HDD
components within commercially available
medium-to-high-function storage systems typically
cost less than 10 percent of the cost of the system.
Customer value has migrated to advanced func-
tions and the integration of these functions within
the system itself. Then, as Figure 3 shows, the cost
of managing storage now dominates the total cost
of a storage system.
24
This means that the value to
the customer of a storage system now resides in its
ability to increase function beyond what is provided
in the bare
HDD
, and specically in its ability to lower
management costs and provide greater assurances
as to the availability of data (e.g., through backup
and replication services). Buyers of storage systems
are now mainly buying the function that is embod-
ied in the software (sometimes called the rmware)
of the storage system. Furthermore, buyers of stor-
age systems are discounting the initial purchase
price in the buying criteria and weighing the total
cost of ownership more heavily.
In the present issue of the IBM Systems Journal, we
have included papers that document the ongoing
evolution of some key storage system technologies.
In the rest of this paper, we rst discuss how these
new capabilities were driven by technological devel-
opments, such as local area networking, and by the
IT
(information technology) requirements for the en-
terprise. Then we show how todays challenges in the
industry are evolving into the challenges of the fu-
ture.
Storage systems come of age: from
components to systems
It has long been recognized that the disk drive alone
cannot provide the range of storage capabilities re-
quired by enterprise systems. The rst storage de-
vices were directly controlled by the
CPU
. Although
some advances did take place in the interim,
System/360* in 1964 was the rst offering of advanced
functions in an external storage control unit.
5
The
key advantage of a control unit (or controller) was
that the
I/O
commands from the
CPU
(sometimes
called the host) were independently translated into
the specic commands necessary to operate the
HDD
(sometimes called the direct access storage device,
or
DASD
), and so the
HDD
device itself could be man-
aged independently and asynchronously from the
CPU
. The control unit included buffering that allowed
Figure 2 Improvement factors for PC technologies: since

1990 storage technology has outpaced both

processor and communication technologies.
1980
1985
1990
1995
2000
2005
2010
2015
10 000
1000
100
10
1
IMPROVEMENT F
ACTOR
YEAR
PROCESSOR
1983@0.33 MIPS
STORAGE
1983@10 MB
COM