NightHawk Sought a Consistent Platform
A Year Later, Cisco Making Serious Headway with Its UCS
When networking king Cisco Systems officially entered the full-service data
center systems market on March
16, 2009, with the launch
of its Unified Computing System, it jumped into a pool of sharks named IBM,
Hewlett-Packard, Dell, Fujitsu and Sun Microsystems (now Oracle).
Plus, it took a major risk in opening a whole new field of business during the most dangerous macroeconomy since the Great Depression in the 1930s.
That investment and risk in time, capital and personnel appear to be paying off, despite the competition and a dour economy: In only one year since it started shipping the network-centric UCS (Unified Computing System) systems in July 2009, Cisco now counts more than 1,000 enterprises running them on a daily basis.
eWEEK also has learned that Cisco has a lot more customers in the pipeline as the world economy continues to repair itself.
The UCS consists of a proprietary Cisco data center architecture, Cisco-made servers, and a set of management software and services based on Intel's quad-core Nehalem Xeon processors. Cisco partners are providing all hardware and software that aren't in the networking realm; storage, though offered first in the UCS partnership by NetApp and EMC, can be supplied by virtually any vendor.
Use case example: NightHawk Radiology Services
One of Cisco's key UCS deployments involves the largest teleradiology provider in the United States, NightHawk Radiology Services. NightHawk, like any institutional IT system user, has made a leap from the familiar confines of an old-school IT system into the new-generation UCS data center; any switch of that nature is never an easy transition.
NightHawk provides digital MRI (magnetic resonance imaging) and CT (computed tomography) services-including 3D images-to about 1,600 hospitals in the United States (26 percent of all hospitals) and employs 144 radiologists scattered around the world.
MRI is primarily a noninvasive medical imaging technique used in radiology to visualize detailed internal structure and limited function of the body. Computed tomography is a medical imaging method employing tomography created by computer processing.
Digital geometry processing is used to generate a three-dimensional image of the inside of an object from a large series of two-dimensional X-ray images taken around a single axis of rotation. MRI and CT produce a volume of data that can be manipulated through a process known as "windowing" so as to demonstrate various bodily structures based on their ability to block the X-ray beam.
NightHawk distributes a huge number of digital imaging files on a daily basis-from hospital to data center, from data center to radiologist, from radiologist back to the data center, and finally back to the hospital.
"We do about 9,000 studies [individual radiology sessions] a day-a little over 3 million a year," NightHawk Vice President of Information Technology Ken Brande told eWEEK.
"A study," Brande said, "is any type of medical image transaction, whether it's an X-ray on a broken bone to a chest-abdomen-pelvis X-ray that shows a progression or regression in growths of a cancer. It's any and all of those things that typically would be offered in a hospital setting for viewing of a patient."
Huge number of digital images processed daily
Nine thousand sessions involving medical images is a ton of data to be transported through secure IT pipelines each day.
NightHawk's procedure works this way: Each hospital has a virtual private network (VPN) tunnel to one of three NightHawk data centers. As the images enter the system, they are immediately distributed to one or more of the 144 radiologists on the NightHawk staff, who are on continuously overlapping shifts that cover the 24/7 spectrum of the business.
Amazingly, despite the glut of images pouring through the UCS, medical determinations based on the digital images are made very quickly.
UCS Executes Two Basic Workflows
"Our average turnaround time-from the time we receive the study to the
time we get the report back to the hospital-is less than 20 minutes,"
Brande told eWEEK. "This is relevant because a big portion of our studies
are for emergent care-there's somebody waiting for the report in order to know
how to treat the patient."
There are two basic workflows involving the UCS system, Brande said.
"One would be the order, or requisition flow; that is a NightHawk application in which the hospital will go in and enter information about the patient," Brande said. "This will tell us the patient's history. That requisition has to be tied to the images for that particular order.
"The second flow is getting the actual images from the hospital to us. We process [the images], link those two things together, then we send that out to our radiologists."
About 80 percent of the radiologists work from remote locations, Brande said. The system can reach them, as long as they have a laptop computer, anywhere in the world. The Cisco UCS makes all this happen smoothly, Brande said.
Company existed on 'ad hoc server rooms'
Before it started deploying the Cisco UCS last winter, NightHawk had for about six years what best can be described as "ad hoc server rooms," Brande said.
With the continued growth of the business, "a lot of what we deployed came out of 'Where will it fit?' as opposed to any specific design," Brande said.
"We ended up with a lot of pockets of infrastructure in business centers around the world. Those rooms were not necessarily built to be hardened data centers. We had power issues, not-always-appropriate cooling-the obvious impact of that type of stuff was that you increase the opportunity for service-impacting outages-things that typically shouldn't become an issue. Manageability was a real problem," Brande said.
NightHawk had 120 servers dispersed in three countries. Upgrading firmware, issuing software patches, moving assets from one place to the next-it was all a serious and costly problem, Brande said.
The first thing NightHawk wanted to do was consolidate all the ad hoc server rooms into a more managed environment.
"We wanted to do that without breaking the bank," Brande said. "And we wanted to free up our resources to be responsive to other challenges in the business."
Those involved the fast-increasing number of medical images to process on a daily basis in addition to improvements in imaging software, which NightHawk was anxious to get installed and running as soon as possible.
How UCS stepped into the picture
That's where the UCS stepped into the (digital) picture.
"[In the past], we had been buying whatever was on the market that best fit the purpose of what we needed," Christopher Smith, manager of data center infrastructure at NightHawk, told eWEEK. "They were fine when they were installed, but inappropriate for when volumes grew."
Multiple vendors and multiple versions of software and firmware caused "a bit of a nightmare keeping track of everything," Smith said. "We were spending far too much time figuring out where everything was, instead of figuring out how to make things better."
NightHawk Sought a Consistent Platform
NightHawk was attracted to Cisco UCS because it was looking for a consistent
platform, Smith said.
"We could put our hardware into it-our blades in general-and we knew we'd be relying a lot on virtualization," Smith said. "We already had a non-trivial investment in virtualization, using the Xen hypervisor-mostly in testing. When I came to start comparing the various hardware solutions, it became clear to me that the UCS had been built from the ground up to be a virtualization powerhouse.
"It has 10Gb Ethernet [networking] throughout, it has extended memory-and that's usually the problem with virtualization, is that it runs out of memory before anything else, particularly with the powerful processes we have today.
"The unified fabric makes things a lot simpler, with less cabling, less infrastructure in general. Performance is excellent, too. Not having to drop Fibre Channel switch modules into every single blade chassis and separate HBA mezzanine cards into all the blades was also quite attractive," Smith said.
Manageability greatly improved
Smith also said the UCS Manager allows him to "manage just about everything you could want to do with the hardware-everything from what VLAN is being trunked to a machine to revisions of firmware to BIOS settings. It just made everything a lot easier."
Most of the alternatives were "a bunch of separate programs from the same vendor glued together in a bit of an ad hoc, patchwork interface," Smith said, "or they really were separate programs from separate vendors that sort of were bummed together in a package, with no real coherence. The UCS was, quite frankly, awesome from that perspective."
NightHawk now has configured eight Cisco UCS blades as VMware ESX servers, each hosting approximately 10 virtual machines, and has plans to double the number of virtual machines in the future, Smith said. NightHawk also uses the Cisco Nexus 1000V Switch, which operates inside the VMware ESX hypervisor to give the network team visibility into the virtual machine environment for more effective management, Smith said.
NetApp provides the scale-out storage for those thousands of images per day.
NightHawk has deployed a full suite of Cisco data center solutions. In addition to UCS and Nexus 5000 data center switches, the company also deployed Cisco Nexus 2000 Fabric Extenders, Cisco MDS storage switches and the Cisco Catalyst 4900 switch as part of the overall infrastructure. These enable NightHawk to deploy a unified fabric in the data center that uses less power and is easier to run, Smith said.
Cisco Services and World Wide Technology supported the project with design, planning and deployment details.
'Vendor lock-in' not a problem here
Does NightHawk worry about being beholden to a single vendor?
"Certainly in a business that has high growth and a need for flexibility, vendor lock-in can be very beneficial," Brande said. "It makes the decision process for which technology to use more streamlined. I'm not having to consider multiple proposals every time we need to deploy a new server or for a new capability."
Lock-in isn't always the correct term, Smith said.
"There are certainly vendors out there-once you start using their stuff-who will go out of their way to make it hard to get off it," Smith said. "With one [trusted] vendor, the hardware decisions become so much easier-the decisions almost make themselves."
Smith said that NightHawk has servers from all the major vendors in its legacy environment and that the Cisco system interacts quite well with them all.
"I wouldn't look at anything in the UCS as being isolating in any way," Brande said. "We can move our virtualization environment off the UCS platform relatively easily. But then we reintroduce a significant number of the operational inefficiencies we had before."