UnityPoint Health is a large multicampus health system serving patients across Iowa, Illinois and Wisconsin and is currently managing two large data centers. Instead of continuing to purchase more hardware-based servers and expanding these data centers, leadership decided to move to a completely virtual server environment. This project was made more urgent by a hardware replacement cycle. We had passed the four year life span of the equipment and were faced with the prospect of replacing not just the 12 servers in our main center, but the mirrored system in a disaster recovery location.

We knew maintenance and power costs would rise as we expanded our operations. We had to grow our application server environment, but more hardware through capital expenditures no longer made sense.

Developing support for what some saw as a “big leap” to virtualization was critical. Internal communications and transparency were key and we worked closely with our PACS Governance Council to gain their consent needed within our IT governance framework.

An important element to achieving the consent was proving that “the big leap” would occur without any gaps in service or performance. Stakeholders needed assurance that the system would continue to meet the internal service level agreements.

At first, we moved cautiously, implementing the virtual environment while still maintaining our physical servers. We made sure all the applications worked in our test system and only then did we change out the first physical server for the first virtual server. From then on, we switched servers one at a time. A week later we were unracking the old servers and getting them ready for removal.

Before switching to virtual servers we were seeing a lot of study sequencing and study acquisition related problems caused by a system operating at max capacity and above. These problems cause excessive administrative overhead for both our system administrators and our vendors’ support team. The move to a virtual environment resulted in much less administration time devoted to correcting application errors that were created by the physical hardware operating at and above capacity. Radiologists and clinicians were able to access and view exams in a faster and more efficient manner and we recorded a 175 percent increase in the ability of an application server to process incoming images.

We improved performance in almost every area on the test system. One of the most impressive discoveries was in the area of disaster recovery. Bringing up our physical server-based mirror site required two hours with a full 45 minutes devoted to failing over applications to the servers. Our virtual disaster recovery system came up in less than three minutes.

Data from Microsoft Performance Monitor demonstrated our servers were typically running at 60 to 70 percent during peak loads. After the switch to the virtual machine, our processing dropped to just 20 percent at peak.

The virtualized environment also provided better network load balancers. With our own network switch within the Citrix NetScaler application delivery controller we had an independent configuration that didn’t have to compete with other users.

The move allowed us to upgrade to Windows 2008 allowing us to achieve a dramatic improvement in performance. These accelerated standards can be attributed not just to an improved operating system, but also to better hardware in the virtualized environment. We went from individual one gigabit network connections to a large virtual machine server farm with shared 30 gigabytes of bandwidth and a ten gigabit mix on the application server network cards.

We saw these improvements in our performance testing throughout the conversion process. Processor, network and disk queues demonstrated that while a large numbers of operations were waiting, on the virtual side, there were no operations queued. With this experience, we started to virtualize other applications to achieve the same kind of performance benefits. These solutions included dictation voice recognition systems and qualitative intelligence and communication systems (QICS); our workflow and communication solution.

During this process we also added a browser-agnostic viewer that allows physicians access from a variety of mobile devices. With the addition of an Enterprise Image Repository we were able to achieve a vendor neutral archive that effectively met the needs of our growing affiliate health systems.

For UnityPoint Health, moving into the unfamiliar territory of creating a virtual environment for imaging and other applications was more than worth it. As the number of scans
has grown from 800,000 to more than 1.2 million and rising, the system has continued to exceed our expectations. The results included both greatly improved performance coupled with a significant reduction in costs.

About the author: Matthew Bishop is Enterprise Solutions Architect at UnityPoint Health. Its network includes 17 UnityPoint Health Hospitals and 15 community network hospitals.