Cancer center SAN ensures on-time treatment

InSite: Lessons from leading users

At the Cancer Therapy & Research Center in San Antonio, Texas, more than 200 patients each day receive radiation therapy on time because of an iSCSI storage-area network and some innovative remote boot technology.

At CTRC, eight devices known as linear accelerators need to deliver radiation doses to CTRC patients every 10 minutes -one linear accelerator per patient.

Cancer Therapy & Research Center CTO Mike Luter says keeping critical components in his data center available at all times is key to the center’s care offerings.

Each accelerator communicates with a Windows NT server, which records and verifies radiation doses. The NT server, a critical part of the center’s network, must stay up and running at all times so patients can get their treatment, says CTO Mike Luter.

Every minute counts

If the server is down for 10 minutes, as many as eight patients might miss life-saving radiation treatments. If a server is down for an hour while network managers rebuilds it, it can set the entire day off -50 patients won’t receive treatments.

In this installation, the server connects to a 3-terabyte EMC Clariion FC4700 disk subsystem, which stores the 100M-byte-sized patient treatment plan generated from the computed tomography imaging scans, magnetic resonance imaging and Positron emission tomography scans.

The FC4700 in turn links into to the Gigabit Ethernet network via dual Cisco SN5428 iSCSI storage routers. The Unix, Linux and Windows servers that run CTRC’s diagnostic applications also connect to the Gigabit Ethernet network and access the data stored on the FC4700S via the iSCSI protocol.

The boot

To keep the server that records and verifies radiation doses up and running, Luter uses software called the Cisco Network Boot, which allows a spare server to load the operating system and initialization processes and step in for the other server if it fails.

Network Boot allows CTRC to put the contents of the server’s system drive, or “booting function,” on the FC4700 instead of the hard disk of the server, so when a server fails, boot information and data is not lost.

Coming together

“All I need to do now is go to the [Dynamic Host Configuration Protocol] server, which has the [media access control] address of that record and verify server, change the address to that of the spare server, boot it, and now it’s the record and verify server,” Luter says.

Using Network Boot serves as an alternative to provisioning another server and lets Luter quickly replace a failed server, so treatments are not held up.

The data on the FC4700 disk array is mirrored using Veritas Software’s Volume Manager for disaster-recovery purposes to a location 22 miles away, where Luter has another FC4700 subsystem and redundant SN5428 storage routers installed.

Fibre Channel costs

Initially Luter planned to link the two SANs with Fibre Channel. He soon found that the cost of the Fibre Channel link was prohibitive and instead decided to adopt iSCSI. Observers have said that mirrorine data using Fibre Channel over IP can cost as much as $2,000 per month, per mile.

“Our clinics and our research park are 22 miles apart,” Luter says. “Its hard to run Fibre Channel 22 miles, even with SAN extension.

“We already had Gigabit Ethernet between the two sites so we could share the connection among voice over IR applications and the iSCSI mirroring,” he adds. “I didn’t have to buy another circuit and didn’t have to put expensive Fibre Channel adapters in my servers to accomplish it.”

Luter’s existing IT staff is in charge of maintaining the iSCSI network.

Copyright Network World Inc. Nov 18, 2003