With the cost of hard drives declining, a single-tier archiving approach is the best solution for many imaging facilities, based on reliability, access, and ease of scalability.
The volume of archived digital images is increasing and will
continue to rise more steeply than film storage volumes did in the
past. Many filmless facilities have been caught off guard by this
increase, which has been stimulated by many factors. The most
significant is investment in new modalities that are digital and
compliant with the Digital Imaging and Communications in Medicine
(DICOM) standard. This investment brings more images into the
filmless network. Computed radiography and digital radiography
technologies are becoming more affordable, so more plain-film
studies are now digital. A huge volume driver is the increase in
images per study from multislice technology. New multislice CT
scanners, for example, can generate as many as 800 to 1000 images
per examination.
Storage requirements also are affected by disaster-recovery
initiatives and state retention mandates. Health Insurance
Portability and Accountability Act (HIPAA) guidelines for disaster
recovery and data storage require the existence of a backup of
imaging data that can be quickly recovered in case of disaster.
This can be accomplished in a number of ways, but they all mean
that a second copy of the data must be stored, which is most easily
accomplished electronically.
THE TRUE COST OF STORAGE
Barbara Dumery
|
Burgeoning imaging data volume presents both cost and technology
challenges. The total cost of long-term ownership is an extremely
important consideration, but is not always easy to calculate. It is
important to look at both initial costs and recurring costs. As a
facility considers implementing a picture archiving and
communications system (PACS), it is difficult to predict future
needs. Therefore, there must be a balance between scalability and
cost containment. If cost were the only consideration, without
regard for future scale, future costs associated with outgrowing
the technology could negate the original return on investment.
There are several factors that contribute to the total cost of
ownership, and that the enterprise should consider as it evaluates
any archiving solution. As the amount of information increases, and
dependence on these data increases, it becomes more important for
the data to be retrieved quickly and easily. Some archiving options
have a lower cost but have slow access times. The time of
radiologists and technologists is valuable. As part of any cost
analysis, data-access times for archived information should be
evaluated.
Although most types of storage media have been on the market for
a long time, their storage formats and capacities are continuously
being improved. This means that the drives that read the media also
change. A 5-year-old tape cartridge may no longer be readable using
current drive technology. The data must then be migrated to the
newer technology, and that can introduce additional costs.
Ever-increasing storage capacity also means increased demands on
the storage setting. More physical space is required, and some
solutions require more space than others. Some storage technologies
also have specific requirements for power, cooling, service access,
security, and so forth. These data-center environmental issues must
be addressed, and associated costs must be considered. As a
facility moves toward becoming a filmless enterprise, its
dependence upon reliable, robust systems increases. The redundancy
and reliability built into the system must strike a balance between
budget constraints and a comfortable degree of risk.
The trend favoring online storage is facilitated by a continuous
decrease in hard-disk cost per MB. In 1998, some analysts correctly
predicted that the price per MB for hard disk storage would drop
dramatically by 2002, based on a 10-year trend leading from $11.54
in 1988 to to 4.3 cents in 1998. Today, the overall average price
per MB continues to decline. For that reason, online hard-disk
archiving has become a viable storage solution for the primary
archiving of images.
EVALUATING AN ARCHIVE SOLUTION
The two basic approaches to archiving are single tier and
multitier; each has benefits.
The traditional approach to archiving has been a multitier
approach in which all images are stored at the deepest archiving
level (usually a jukebox with tape or optical disks). This was an
inexpensive storage method, but at the expense of immediate access.
To improve performance, only the images used most frequently were
stored on high-cost, fast-access media, usually a redundant array
of inexpensive disks (RAID) or direct attached storage (DAS). Those
less frequently accessed were shifted to lower-cost,
slower-retrieval media. Images from the deep archive were moved
back to the intermediate and online levels (usually RAID) if they
were needed more quickly.
As the cost of hard-disk storage declines, there is an
increasing trend toward a single-tier archiving approach. All the
images are stored on a single medium that can be accessed very
quickly (this is usually RAID, or network storage). A redundant
backup copy of the images is then stored on another, less expensive
medium, such as tape or modular optical disk (MOD). In this
approach, online storage is increased incrementally as volume
grows.
Network storage is simply a system architecture whereby a
server, which is dedicated to nothing more than file storage and
sharing, sits on the network. Data stored on the network are made
available directly to clients on the network. Two methods of
archiving fall into this category: network attached storage (NAS)
and storage area network (SAN). These network storage devices do
not provide any of the activities that a server in a
server-centered system typically provides, such as image
management, compression, or routing. Network storage allows more
hard-disk storage space to be added to a server-centered network
without shutting the host server down for maintenance and upgrades.
Unlike DAS, network storage devices do not need to be connected
directly to the server itself, but can exist anywhere on a
limited-area network.
Whether a single-tier or multitier approach offers the lower
long-term cost of ownership really depends on a number of
factors.
COST OF OWNERSHIP
For fast access to all images with high scalability, the
single-tier approach is recommended. Since all images are online,
all studies, both current and historical, are immediately available
without the need for prefetching. The end user is then empowered to
select and view the prior studies that are relevant. The major
benefit of the single-tier approach, especially when using NAS or
SAN, is that the storage capacity can be increased as needed,
without interruption to systems and work flow. The initial
investment consists only of the infrastructure and storage needed
for 1 to 2 years. As time passes and storage requirements grow,
more storage can be added easily. Rather than making a premium
investment in total hard-drive capacity today, the organization can
take advantage of the ever-decreasing cost of drives, adding more
as prices decline in the future. For smaller enterprises, NAS or
SAN can be cost prohibitive, and a multitier approach might be more
suitable.
The cost benefit of a multitier approach sometimes outweighs the
decrease in scalability and performance. In environments where
little growth is expected, these latter factors are less
significant. Both scalability and performance can be achieved, but
there are issues to consider. First, to achieve good performance
using a multitier approach, prefetching is a must. Algorithms must
be developed to determine when (and which) prior studies need to be
migrated from the deep archive to online storage. This introduces
complexity and limitations to the number of prior studies that may
be available to the end user. To achieve scalability in pure
jukebox environments, additional jukeboxes can be added. Because
the cost of housing and robotics can be high, it is usually most
cost effective to plan for a jukebox large enough to accommodate 5
years' imaging volume. Reliability is a critical factor in any
archiving solution. If the system is down, clinicians cannot get to
images, and radiologists cannot do their jobs. Both single-tier and
multitier systems can be configured for reliability. Single-tier
systems are inherently more reliable than multitier systems by
virtue of the fact that data are stored redundantly on hard drives.
In a multitier approach, buyers should consider only solutions that
offer highly reliable removable media. There are many vendors with
jukebox solutions, some more reliable than others. Jukeboxes have
more moving parts and robotics. The more moving parts, the more a
system is prone to failure. Although some vendors' solutions are
self calibrating, many offer jukeboxes requiring calibration.
Organizations should be sure to ask the PACS vendor what jukebox
technology is used and what reliability record it has shown.
Backup is a critical component of any disaster-recovery plan.
Redundancy, both in system architecture and data, is a major
consideration, especially with the advent of HIPAA guidelines.
With respect to system architecture, critical components within
a server must be redundant. Beyond that, a failover server can be
employed in case the primary server fails. Failover is a mechanism
by which an operation automatically switches to a standby database,
server, or network when the primary device fails. Ideally, these
two servers should be in physically different locations in case of
a disaster. Fully automated failover is not a prerequisite for an
effective system in medical imaging, but there must be a mechanism
in place to avoid downtime. Data redundancy can be achieved using a
copy of data on removable media. The media should then be stored in
a physically different location. Automated forms of backup can be
considered, such as a redundant jukebox or NAS. Backup copies of
the database should also be maintained.
It is necessary to strike a balance between cost and the level
of risk mitigation that the organization is willing to undertake.
As the organization makes decisions on its archiving solution,
consider that it will need access to imaging data 7 years from now
(or longer). Bearing the costs of technology compliance over time
can be overwhelming. When thinking about any storage medium, one
should ask what the expected future advances will be, whether those
will be backward compatible, and how easily the data can be
migrated from the old medium to newer media. If the organization
chooses not to migrate from old drive technology, however, it will
not gain performance. Servicing old removable media drives can also
be a costly challenge. With a single-tier approach, media/drive
compatibility is a less important issue. With hard-disk drive
technology, the drive and disk are inseparable (the drive can
always read the disk).
Whenever possible, simplicity and minimization of system
administration are needed. A system is more cost effective when
monitoring and administration are easy and convenient, with both
local and remote access. Notification and alerts are also critical
because they can head off potential problems. These goals can be
met with both single-tier and multitier approaches, but the cost of
ownership is less with a higher level of administrative simplicity.
The physical space of an archive is important to consider.
Multitier solutions rely heavily on physical components that
require considerably more floor space than hard-drive based
single-tier solutions. NAS packs 2.25 times more storage than MOD.
Since the cost space is real, this expense should also be factored
into the long-term cost-of-ownership analysis. For single-tier
system, the initial cost is the server and NAS or SAN
infrastructure. For smaller and mid-sized facilities, SAN is still
cost-prohibitive because of its infrastructure costs. The recurring
costs in a single-tier approach are service and additional storage
increments. With the multitier approach,the initial cost is often
higher than for NAS. These costs include a server, some form of
short-term storage (usually RAID), and the long-term archive
(usually a jukebox). The recurring costs would also be service and
the cost of additional media or more jukeboxes.
CONCLUSION
For most imaging enterprises, a single-tier archiving approach
is the best solution. With the cost of hard drives declining, NAS
is a very feasible solution today. It is highly reliable, it offers
immediate access to all examinations, and it easily scales as
imaging volume grows. Best of all, media-obsolescence challenges
need not be a concern.
For backup storage, a MOD or tape jukebox can be implemented.
This requires a smaller investment as it will only be used for a
redundant copy of the data. There is no need to keep the data
online and available. If further system redundancy is desired,
multiple servers should be considered.
If the enterprise already has a SAN in place for other purposes,
take advantage of it. Otherwise, SAN is neither affordable nor is
it necessary for an effective PACS.
The multitier approach still has its merits for smaller
enterprises, but with a detailed long-term cost-of-ownership
analysis, NAS will probably still come out on top as the solution
of choice for many imaging facilities.
Barbara Dumery is senior product manager, eMed Technologies, Lexington, Mass.