For most organizations, Big Data is still a future technology. In medicine, and in particular for major medical centers that combine advanced treatment of genetic diseases such as cancer with research, it is a present-day reality. The grainy grey-scale snapshots produced by medical imaging equipment just a few years ago are now full color, three-dimensional, high-resolution images or even videos. Patient data that a decade ago would have been measured in kilobytes, now can include megabytes of genetic and treatment data. High speed, dependable access to this data can literally be a matter of life or death.

This was the challenge that Tahir Ali faced when he came to City of Hope National Medical Center in the summer of 2013. “At the time we had Dell Compellent as our enterprise block storage solution, and it was not working as well as everybody hoped for,” he said. “The levels of IOPS that we were expecting were much higher than what the existing infrastructure could deliver.”
Ali, who is senior manager of infrastructure engineering at the cancer center and who appeared on theCUBE from IBM Edge 2014, had background in Apache, IBM and EMC systems. He went looking for a solution that would provide high performance in a very high input/output operations per second (IOPS) environment, with high levels of dependability and automation in an affordable package. After considering technology from NetApp, Inc.
and EMC, he chose IBM XIV for the core system. XIV, he said, provides storage performance that is just below Tier 1 – he calls it Tier 1.25 – and is extremely dependable and less expensive than EMC VMAX.

At that time IBM was just bringing its initial FlashSystem 820 all-flash array to the market, and Ali immediately saw the huge performance advantage of the flash system, and in particular its very high IOPS. He had City of Hope add a fully loaded 820 as a tier 0 front-end to the XIV. When the 840 came out, he upgraded to that model, giving the hospital 80 Tbytes of flash backed by 320 Tbytes of high-speed disk on the XIV.

Most companies with similar physical infrastructures mirror the data on flash to disk. This gives very high levels of data redundancy but cuts the available storage by 80 Tbytes and means that tiering cannot be automated with IBM’s SAN Volume Controller (SVC) Easy Tier.

Automated tiering

“I looked at the flash array’s performance and was satisfied that it was highly dependable and that mirroring was not necessary,” Ali said. So he instead combined the flash array and XIV into a single logical array, virtualized it with SVC and let Easy Tier automatically move the data most in demand to the flash system and the cooler data onto the XIV. This eliminates a great deal of manual work and ensures that City of Hope gets the maximum benefit from its flash investment.

Mirroring did play an important role, however. When the IBM array was installed, the initial issue was moving a large volume of data off the Compellent array with zero downtime. Ali’s answer was to take advantage of SVC’s ability to work with third-party arrays to mirror the data between the two systems. When all the data was mirrored onto the new array, he made that the primary and shut down the Dell system and had it removed from the data center floor.

That agnostic nature of SVC is one of the reasons Ali chose it for virtualizing his storage. Another reason was data redundancy. “I personally don’t think there is any technology out there for block-to-block replication that is better or more sophisticated than SVC.”

Performance is the key

Flash storage, famously, has a raw read/write performance that is an order-of-magnitude faster than high-speed disk. It is one thing to read or hear this and another to experience it, as anyone who has moved from a disk-based laptop to an all-flash tablet knows. At City of Hope, the difference astounded the oncologists, nurses and medical researchers who are the system’s primary end-users. “Sometimes a user would think something was broken because the processes they started happened so fast they didn’t realize it had started,” Ali said. Doctors who were used to starting a process and going to get a cup of coffee while huge data files were transferred or analyzed had to recalibrate their expectations. Sometimes answers came almost before the questions were finished.

The IBM flash array is both faster and less expensive than the competition, he said, partly because IBM does not buy commodity SSDs. Instead, it buys flash chips directly from Toshiba and builds its own SSDs optimized for performance, efficiency and reliability, says IBM Director of Information Technology Michael Kuhn.

Internally, the entire data center is virtualized and IT services are delivered in a private cloud. That frees the staff from putting out fires manually to focus on innovative technologies. That means City of Hope gets more value from its systems engineers. “We have to be agile, and IBM has made us agile,” Ali says. And the service from IBM, when something does go wrong, is extremely good. When City of Hope replaced its original FlashSystem 820 with the higher capacity 840, for instance, IBM handled the data migration.

“We get very positive feedback,” says Ali. “What used to take hours, for instance in our ERP, now is completed in minutes. That has made a real change in how our doctors and researchers can work.”

Courtesy City of Hope National Medical Center