What the financial industry knows about IT performance… and healthcare should

In the decade prior to 2008, there were many noticeable differences between the two types of servers, with many larger hospitals and hospital systems preferring the then-superior performance and control that Linux servers offered. However, Microsoft was continuing to close the gap leading up to 2008.

With the release of Microsoft Windows Server 2008, Microsoft eliminated the gaps in performance, reliability and security. Today, 7 years later, there are no differences in performance as confirmed in independent tests. Many health IT professionals readily recognize the additional cost advantages Microsoft Windows servers offer and the greater availability of qualified IT professionals to staff their growing departments. In fact, all healthcare facilities today, regardless of size, utilize Microsoft servers in one form or another.

The hospitals that continue to utilize Unix solutions for their interoperability/interfacing activities do so primarily due to staff resistance to change, staff preferences for applications that require significant programming and coding, and other biases that can hinder healthcare organizations in various ways, notably in additional costs and scalabilty.

Having a robust operating system that can handle the load of an integration engine is required for the data demands of the healthcare industry. Large providers are known to utilize a single server to run hundreds of interfaces and millions of daily messages. A single Corepoint Health server, for example, can handle more than 10 million health data messages per day, concentrated at the peak hours of the day.

Real-world testing

Tests have been performed in other industries that compare Microsoft Windows and Unix solutions under similar circumstances and the results consistently show that Windows performs equal to or greater than Linux configurations. The added benefits of Windows (affordability, security, professional familiarity) typically tilt the scale in favor of Windows servers. One such comparison at SAP is highlighted on the next page.

Scenario 1

To compare the load-handling of Unix solutions versus Windows, SAP Software and Solutions compared the use of its global ERP application running on Windows with the same application running on SuSE Linux. Because SAP software applications are some of the most demanding across all industries worldwide, the study offers a solid benchmark for how the two operating systems used in healthcare will perform under similar circumstances.

The SAP test¹ consisted of over 3 million line items processed in a given hour with over 30,000 users. The following results show that the Windows Server 2008 processed approximately twice the number of items processed on the Linux server. While the configuration of the servers are not identical, after normalizing the test for processing power, the two servers perform practically identical in an extremely load-intensive scenario.

The numbers of messages processed through interfaces on healthcare servers will continue to rise as the industry moves from a fee-for-service model to the new quality care model, as outlined in the Affordable Care Act.

The need for patient data for population health management and data analytics is placing additional pressure on integration engine vendors to continue to scale their products to keep up with the demands.

The finance industry is an industry of comparison that moves data from entity to entity in large scale similar to what is being expected in healthcare. How do the data demands of the finance industry compare with healthcare IT?

Established names such as Fiserv, NASDAQ, and SunGard are three financial corporations that process an enormous amount of data on Microsoft servers that can serve as examples for healthcare.

Scenario 2

Fiserv

Fiserv is an IT services provider to the finance industry. A leading bank asked Fiserv to design a system that could store more than 65 terabytes of transaction data so customers could look up five years of data (an action that may be utilized in healthcare for data analytics). Scalability was a key concern based on the need to transact large amounts of data across simultaneous users.

To prepare for the anticipated volumes, Fiserv tested² a transaction load of more than 155 million records with 3,000 users using SQL Server 2008 and the .NET Framework 3.5. The original estimation was that the solution would be able to process 500,000 records a minute. After testing, Fiserv found that the solution ended up processing 1.37 million records a minute—almost 3x the original estimate.

Scenario 3

NASDAQ

NASDAQ OMX owns and operates the NASDAQ Stock Market. When markets open, the company processes more than 1 million messages per second. The U.S. data archive alone handles billions of transactions per day and stores multiple petabytes of data.

Much like “big data” applications in healthcare that are expected to be used to manage specific populations of patients and provide key insights for improving clinical decision making, financial traders need insight into current and historical data to make decisions. With data archives continuing to grow, NASDAQ OMX recognized that its existing database solutions couldn’t support projected long-term growth and that it was time to scale up to prepare for the future.

NASDAQ OMX implemented SQL Server 2014 software in its in-memory columnstore to keep its massive data volumes manageable. Data compression alone saved 500 TB of disk space. Reworked queries that previously took a day or two ran in only minutes.

The report³ quotes NASDAQ OMX’s Director of Database Structures Stanley Orlowski: “SQL Server 2014 is a game changer for us. Its in-memory columnstore, advanced statistics, and cardinality estimation are the tools we need to manage our very large databases for the long term.”

Also important to NASDAQ OMX—just as it is vitally important in healthcare—was the ability to encrypt data. Integrated encryption allowed for scalability while reducing the need to swap space on disks used by third-party data encryption tools.

Scenario 4

SunGard

SunGard provides software and technology services for the financial industry. A large bank wanted to ensure that its SunGard trading and position management solution could handle an anticipated surge in transactions and data. This ability to scale was a requirement just to help their traders stay competitive in the financial marketplace.

To scale out its solution architecture with rapidly increasing volumes of transactions and data, the bank migrated from Sybase to a Microsoft SQL Server 2008 solution. After optimization of the new architecture, throughput was measured at more than 160,000 message inserts per second for the exchange trade performance instead of the more typical 30,000 message inserts per second.

This gave the bank confidence that it could meet future challenges with its trading and position management solution and immediately increased throughput by 40 percent.

Using In-Memory OLTP in SQL Server 2014, SunGard realized twice the performance power during testing. Once implemented fully, the solution provides their customers more trading time and reduces operating costs.

In the report,⁴ SunGard’s director of Adaptiv Labs is quoted saying that significant cost savings will result: “We expect to see some reduced CPU activity once the solution is rolled out, and, as a result, some of our servers might not need to be upgraded. Because we host our applications on more than 1,000 servers, we could significantly reduce costs without updating the hardware we already have in place.”