Cluster Monitoring returns a negative value for system load in Windows

When running Confluence Data Center on Windows Server 2012, the system load returns a negative value (-1). When we divide this by the number of cores, the number is -0.25, or -25%, which can be a bit confusing for administrators.

This stack overflow question suggests that it might be possible to use com.sun.management.OperatingSystemMXBean.getSystemCpuLoad() (or getProcessCpuLoad() instead of java.lang.management.OperatingSystemMXBean.getSystemLoadAverage() - but the getSystemLoadAverage() method isn't present in the com.sun package

This blog posts also has a little more information we might be able use to make this a bit less rough on Windows.