Storage: Deciding How Much is Right

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon

Storage, which is crucial to getting the most out of PACS, is often peripheral to the medical professionals accessing it. Their main concern is reliability and quick availability. IT achieves this by planning for, implementing and managing a scalable (and hopefully flexible) storage architecture. But when you factor in exploding data demands, declining storage costs, ever-changing technology and software, robust disaster recovery, business continuity plans and obsolescence the story gets a little more interesting.

Every PACS has a personal element because each archive is planned for, configured and built to meet the specific requirements of the healthcare provider's medical imaging needs. Storage approaches are as varied as the hospitals that implement them. What's certain is that each year brings forth a newer and more sophisticated storage technology with greater capacity and capabilities - especially with the advent of networked storage. The shift has peaked IT's interest in centrally managing storage architectures in order to allocate storage as needed, where needed.

Limited budgets in the past hindered this storage evolution for hospitals. But the good news for administration and IT is budgets for healthcare IT are growing, according to recent studies. Some 73 percent of hospital CFOs indicated plans to purchase digital radiology systems, while 62 percent (of the 460 surveyed) said they expect to buy IT systems over the next five years, according to a survey earlier this year from the Healthcare Financial Management Association.

Since IT is revered as raising clinical quality and operational productivity, hospitals are more concerned about when they are getting a PACS, not why. The archive is fundamental and patients have the propensity to accumulate a trillion bytes of data over a lifetime, so now more than ever it's critical for healthcare providers to create flexible storage solutions.

When it comes to flexible solutions, IT has learned it is not always about replacing the hardware, but about adding intelligent archive software. "The hardware component of the IT budget is not necessarily growing," says Patrick Boyle, business unit manager of IBM Healthcare & Life Sciences. "The growth has really been in the software and services side of the IT budget within hospitals, such as application software [PACS, laboratory information systems and cardiology information systems], middleware and system management tools."


The move toward new technology is more evolutionary than revolutionary and storage software is the next step in the process. Hospitals should first deploy an optimal archive that meets their PACS needs - which may include leveraging existing IT infrastructures - and then look at software technologies that help improve overall management.

AnMed Health, an integrated delivery system in Anderson, S.C., is looking at implementing storage virtualization software once long-term and short-term PACS storage is complete. "In 1998, we made a decision to put in major new clinical systems," says Darrell Hickman, CIO/director of IS. "Up to that point, we had a few, but mostly what we had was for administration. We used the computer to bill patients, now we were using the computer to treat patients." For the computer to be a routine part of clinical treatment processes, the hospital needed a storage infrastructure supported by an unfailing disaster recovery plan.

AnMed installed IBM's DS6000 and DS8000 [formerly code-named the Shark] mid-range to enterprise-class storage technology that replicates data. With clinical systems storing data on the Shark, disaster recovery decreased from 38 hours to less than 45 minutes, says Hickman. When it came to PACS, IT wanted something similar and selected IBM's DS4000 (formerly FAStT) for its long-term archive, an entry-level to mid-range storage technology similar to the Shark. Two units are installed for disaster recovery purposes - one unit at the main medical center and another at a remote outpatient clinic. Short-term storage is on directly attached RAID, but there are imminent plans to move the archive onto the DS4000. Nearly 6TB will be allocated for short-term storage.

"Concurrently with that, we want to move the PACS database from locally attached RAID onto the Shark," says Hickman. "That, too, will be replicated at both the main medical center and the outpatient health campus." Access time to clinical priors is anticipated to reduce significantly.

Storage is not the driver, it's the tool in the decision-making process. "What drives your thought-process are the problems that you need to solve," he says. "You start by needing a PACS, not that you need storage. Because you need PACS, then you begin to ask how you implement PACS and what is the most effective way to do that?" 


Sometimes healthcare providers have to make changes to their storage platform to construct a more flexible storage infrastructure. The Regional Cancer Center (RCC) in Eerie, Pa., began re-thinking its direct attached storage (DAS) system as strategic plans emerged - including PET/CT imaging - that required increasing IT support.

RCC is an outpatient facility that provides more than a half million services per year at five locations serving three states. Treatment specialties include medical oncology, radiation oncology and hematology. RCC sees more than 2,000 new patients annually. Nelson Puello, manager of IT, explains that the EMR system from IMPAC Medical System acts like a PACS. "Images used for treatment are stored in the database of the IMPAC system," he says. "[The database] has been growing exponentially since we acquired a new [multislice] scanner. The system has quadrupled the number of slices that we can acquire and our storage needs have more than doubled."

IT estimated that its storage needs would grow about 44 percent over the next three years. The question arose as to how IT could implement a solution that was scalable and provide storage that was independent from the systems that use it. "The problem with DAS is that if I have a server with X amount of storage, and that server only uses 25 percent of that storage, then 75 percent of remaining storage can't be shared with another application," says Puello. "We needed a solution that would allow us to add storage as we needed, and that it was completely independent from the applications that use them."

RCC maximized prior investments by going with storage technology just beginning to hit the healthcare market, an IP SAN from Network Appliance running on iSCSI (internet small computer system interface) standards. iSCSI-based technology works over standard, low-cost Ethernet networks and avoids the use of more expensive fiber channel communications pipelines - a "boutique" technology, coins David Dale, storage industry evangelist for NetApp. "Healthcare usually has really modest IT staffing levels, much more modest than other segments," says Dale. "So they are looking for something that does not impose higher administrative overhead on their already sort of stretched resources." Managing the IP SAN at RCC only requires the resources of two people.


Healthcare, like many other industries, is  moving to networked storage technology that offers greater manageability and flexibility over enterprise-wide infrastructures. Wake Forrest Baptist Medical Center implemented an enterprise SAN after trying to manage separate storage silos for each of its large departments.

Using a multi-tiered approach for radiology, their front-line storage is EMC Corp.'s Symmetrix, a high performing disc array that stores electronic images for eight months to one year. The second tier, EMC's Centera, is a fixed-content platform that holds onto an image for up to two years. StorageTek's PowderHorn 9310 tape library is the last storage silo that keeps the image, this time for as long as it needs to be maintained.

Bob Massengill, manager of technical services, says the process is facilitated using StorageTek's Application Storage Manager software. "Although our storage arena is located on three different platforms and much of the data are used for image applications, the SAN also is the primary [centralized] storage for many of our applications and departmental data," says Massengill. "We currently have about 100 different applications and departmental systems throughout our organization using our SAN."

Prior to the SAN, each department was storing its data on internal type storage devices. Plagued by high costs, management was a heavy burden. "Departments were not growing out their servers, they were outgrowing the storage piece of it," explains Massengill. "But we could not upgrade just that particular storage, we had to do early replacements of the entire server, which was about every 18 months." With the SAN, Wake Forrest extended the life of the server to 36 and 48 months and put management into the hands of IT.

When figuring out your storage needs, study each application carefully, advises Massengill, particularly when it comes to PACS. "Why would data be stored on 10 cents a MG media that may not ever be accessed when you can put it on 1/2 cent a MG media," poses Massengill. "I think people really need to understand the technology out there today, what costs are associated with that technology and make sure that they understand each application and where the data from each application really need to sit."


Cedars-Sinai Medical Center in Los Angeles is another hospital that realized the time was right to consolidate its storage. The hospital has four PACS, one for each department: Imaging, the Vascular Lab, Echocardiography and Cardiac Catheterization. "There have been separate archives associated with each PACS because back in 2000, most PACS vendors sold systems that were predominantly standalone, turnkey solutions," says David Brown, manager of PACS/RIS at the S. Mark Taper Foundation Imaging Center at Cedars-Sinai.

Since that time, the storage for each of the systems has quite naturally filled up.  The Imaging PACS archive consists of three StorageTek 9840 tape libraries with a total capacity of 50TB of online data. The libraries are projected to reach capacity in early 2005. Camtronics PACS (for echo) was installed with two Rorke Data AIT-2 tape libraries with a total capacity of 4 TB. The HeartLab PACS for cath lab has a DVD jukebox archive, and the McKesson PACS in the Vascular Lab stores on an AIT-2 tape library. All are nearing capacity.

As the hospital approved the purchase of additional archives for the new radiology PACS (Kodak DirectView System 5), Brown knew it was time to start building a PACS architecture where the images are stored on a common storage platform and managed centrally. The costly alternative was to continue purchasing separate storage systems for each departmental PACS.

"Today, PACS has matured to the point where everybody recognizes that you need to have more of an enterprise approach," explains Brown. "Vendors are looking at the entire enterprise and know that they need to be able to integrate with existing resources. For example, we have had an existing IBM SAN here for a long time and we are building it up."

"If I have everything archiving to our SAN, I will just need to buy IBM RAID in chunks to keep each of the PACS' databases online for immediate access for the physicians," adds Brown. "Or we could grow our IBM tape archives so that two to three years of data are online on RAID, and then the data are seamlessly migrated to an IBM 3584 tape library connected to the SAN."

Brown also plans on using Kodak's VIParchive software EIM as the middleware that receives the data from all of Cedars PACS and archives them according to a rules-based storage policy. EIM supports storage management policies based on information found within the DICOM header of each file, allowing for data migration policies based on clinical information.  According to Brown, "This will allow me to keep pediatric and mammography studies archived longer than other studies, simply by writing an appropriate storage policy."


Reminiscing on older generations of PACS storage, quite a bit has changed. It's no wonder that IT may feel as if it's chasing its tale when keeping up to date with plummeting storage costs and next-generation technology. And even though predicting capacity can be calculated per modality, per study, estimating data growth leaves room for error.

Inevitable data growth and obsolescence both factored into the storage decision Jim Bates, director of radiology at Naples Community Hospital and North Collier Hospital, a part of NCH Healthcare System in Naples, Fla., made when it came to PACS.
"Storage is the most important part of your PACS, if you fail in that area you fail in your PACS," says Bates. "Department leaders, PACS coordinators and administrators are many times frustrated that they have to go back to [administration] every year and ask again for financing for more storage or different storage. Most capital committees look at PACS as an equipment purchase, but if you underestimate your needs, you will be back in front of the capital committee many times."

An ASP (application service provider) model from InSiteOne to PACS addressed IT's concerns, says Bates. This pay-as-you-go-archive has no additional up-front costs. In fact, getting budget approval was the easiest part of the PACS project, says Bates.

"You don't have to actually know how many images you are going to produce next year, or if you fall short this year, you still pay for it as you create the images," explain Bates. "The cost was less than the film costs for the images, which was very appealing to administration. They wanted the film costs out. I showed them a way that they could pay less than what they were paying for film, not come out of capital budget for hardware, it did not require any up-keep, FTEs or additional real estate."

Network Storage Solutions: A Snapshot

Data storage has evolved from a back-office concern, or something that is attached to a single server or individual computer into an executive issue that requires a top-down strategic plan. HIPAA, disaster recovery, patient care, legal and business continuity all require a coordinated approach to data access, security and storage. A hospital or organization's size will ultimately determine the best strategy. At this time, smaller facilities may not be able to afford network storage solutions, however this will change.

  • Use an enterprise approach. Start thinking enterprise for management of electronic medical images, not separate storage systems for each application in cardiology, radiology and pathology. Also, be sure to invest in gigabit networks - 10 GB for the backbone, and 1 GB to the modalities and clients.
  • Storage strategies. Online storage: Immediately available; millisecond access times; magnetic hard drives (RAID); have one year's worth of online storage (100,000 radiological procedures generate approximately 5 terabytes of data annually). Clinical priors archive: Immediately available; allows rapid retrieval; uses RAID. Disaster recovery archive: Off-line; uses removable media - with a major consideration being cost and reliability of the media. It uses hierarchical storage management (HSM) software to manage various storage options as one unit.


  • DAS (direct access storage) is a general term for magnetic disk storage devices attached to computers or workstations.
  • RAID (redundant array of independent or inexpensive disks) is a way of storing the same data in different places (thus, redundantly) on multiple hard disks. RAID has several levels.
  • NAS (network-attached storage) is hard-disk storage set up with its own network address. It is dedicated to nothing more than file sharing.
  • SAN (storage area network) is a high-speed subnetwork of shared storage devices, but may also extend to remote locations for backup and archival storage, using wide area network carrier technologies such as ATM or SONET.
  • Fibre Channel is a technology for transmitting data between computer devices at data rates of up to 1 or 2 Gbps (and 10 Gbps in the near future).
  • iSCSI (Internet Small Computer System Interface) is an IP-based storage networking standard for linking data storage facilities over intranets and to manage storage over long distances.
  • CAS (content addressed storage) is for data that are not intended to be changed once stored, a.k.a. fixed content data. CAS allows users to manage massive amounts of fixed content (original studies in CT, MRI, archived e-mails and electronic documents) and to be sure of its authenticity and integrity.

Virtualization Software: Meeting Storage Challenges

In the past, applications in hospital departments have been closely tied to their respective servers and storage resources. The problem was that once an application's archive reached capacity, IT cannot turn to the other departments for additional storage. The coping mechanism: storage virtualization software.

"Studies indicate that on average only 40 to 50 percent of the available storage capacity in a hospital is actually used," says Patrick Boyle, business unit executive of IBM Healthcare and Life Sciences Solutions. "What's key to virtualization is to drive utilization of available storage resources to the 80 to 90 percent range."

According to EMC's director of technology analysis Ken Steinhardt, virtualization is having one aspect of an IT system fooled into thinking that it is seeing something it recognizes, when what is actually being presented is really something different than it thinks. Sound confusing? Basically, storage virtualization is the ability to separate software business applications from the specific hardware device. It is the pooling of physical storage from multiple network storage devices - from multiple vendors - into what appears to be a single storage unit that is centrally managed.

What does this mean for administration or IT? Steinhardt explains best/worst case scenarios: "Best case - IT is able to use existing servers and storage devices and only add on additional software. It can integrate it with existing management tools and IT can continue using its existing functional management software. Worst case - IT will have to learn new management tools without integrating into the existing ones, replace existing functional software and implement new hardware components."

To avoid the worst case scenario, Steinhardt suggests considering the following:

  • Is this something that can integrate with and work well with my existing environment?
  • Is this something that will work and play with the evolving industry standards for virtualization of storage networks?
  • Is this something that will give me flexible choice of different vendors and enable me to choose technology at any given point in time without being locked into individuals' proprietary software systems?