Software to detect if a Virtual Machine is busy:

Introduction:

Servers are ubiquitous and a private Cloud provider has hundreds of servers in their inventory. Most of these servers fall under one of the following categories: Initialized, Active, Suspended, and Terminated. A server is generally in the initialized state it is being imaged and prepared for commissioning. A server is active when it is being used frequently by the user. A server is in the suspended state when there is confirmation that the user is no longer using it actively. A server is the terminated state when the server is ready to be reclaimed or repurposed.  Periodically a cloud provider may want to reclaim resources such as servers. In such cases, it is important for the cloud provider mark the servers as suspended before they can be terminated. Going through several hundreds of servers and marking them as suspended only on user confirmation is not very productive. Instead if there were a way to detect whether the servers are active or not, then we could even be more judicious of which users to target for confirmation.

Problem definition: How do we know whether a server is active or not ?

Solution: Servers come in many different OS versions and flavors. A script that might work in one may not work in another. Therefore, we may need to target each OS family with the nuances and metrics that they allow.

On Windows, we have a lot of instrumentation builtin. They came in many different forms such as WMI, perfmon counters, MOM and tracing and logging. There is probably very little need to write sophisticated code to figure out any particular component or system. That said, a Windows system task viewer may show varied level of peaks and valleys of activities even without user input.

On Macintosh, we have plenty of dscripts stashed away under the hood in Automations. In fact, a Mac OS can reliably say when the system is not doing much activity.

On Ubuntu, we have several different counters typically under the proc folder. In fact top and uptime commands often give metrics that give an idea of how active a VM was. The /proc/loadavg has the first three fields as load averages that give an idea of the number of jobs in the run queue. Load average is different from waiting for cpu. Generally if you have 1 process running at 100%, then the load average can be expected to be approach 1.

Conclusion: Periodic cron jobs may collect statistics over time that can reflect the usage of a VM server. Given this visibility, different orchestration frameworks and program may be able to use this information for several purposes such as monitoring, conservation etc.

#codingexercise

int FindSmallestWindow(List<int> haystack, List<int> needle)

{
// parameter validation and assume no duplicates in haystack or needle.

 var indexes= new SortedList();

for (int I =0; I <haystack.Length; I++)

     indexes.Add( haystack[i], i);

Array.sort(needle);

// indexes and needle both sorted

now we can use the sorted sequence and index arithmetic linearly
int min = haystack.Length - 1;
int max =  0;
int range = 0;
for (int i = 0; i < needle.length; i++)
{
int index = indexes[needle[i]];
if (index < min) min = index;
if (index > max) max = index;
}
if (min > max) {int temp = min; min = max; max = temp;}
int range = max - min + 1;
return range;

}