We've recently made several changes to X-Link to improve message delivery times and reduce CPU usage. Some changes are to all editions while some are only to the Enterprise Edition. To help customers understand more about X-Link, we've added a set of performance visualization to all editions to provide insight into X-Link operations. The intent is to show when additional hardware and/or an upgrade to the Service or Enterprise Edition could improve Message Delivery Times and/or CPU Usage.
The Message Delivery Time Visualization Advanced Help
To access the Message Delivery Time Visualization, just press the Delivery Times button in the Visualizations box on the lower left of the X-Link Dashboard as shown below:
A detail box will display and may say Please wait while graph is rendering. A graph, like the one below, should then appear.
This graph represents all messages in all linkages for the date and time selected.
Only delivery information thru yesterday is possible to be displayed. This is due to the way X-Link during it's midnight clean up process, calculates the message deliver times for that prior date.
At this point, you can choose another date and/or choose a specific hour of a particular day. We choose 7 pm hour as it looks like there was a lot of activity at around 7:30 pm followed by light traffic. It is important to note that this graph represents all messages transferred by all linkages. Thus a batch on one linkage may account for the peak at 7 pm hour while the level light traffic after the peak may be from another linkage's real time transfers with normal user updating. In this particular case, that is true, two linkages account for the data displayed. The 1st linkage was connected to an internet based system, while the transfers from about 8 pm on were local data base accesses. Since each endpoint (database, internet server, etc) is different, the "normal" delivery times for each system is different and can be significantly different.
This graph represents the variety of delivery times, from average (dark green line in center), to maximum (red line at top), to minimum (black line at bottom), but it also shows the standard deviation, or distribution of message delivery times (green area around the average). The standard deviation shows us the group of messages that represent the majority of messages sent. For example, the data at 7:28 pm (19:28) shows a message took over 200 seconds, while the majority of messages that minute were sent faster than the minute before and after 7:28 pm, as can be seen by the lower average and standard deviations. This could be explained by a locked patient that X-Link could not update for over 3 minutes, or any number of other reasons, good or bad.
Pointing the cursor at the data in the graph will provide a pop up tip showing the underlying data represented in the graph.
This graph also has an alternate view that shows the break down of the average delivery time between the amount of time to transfer the message verses the amount of time X-Link waited for resources, like data base responses, internet based API servers, hard disks, network lag, CPU contention, and other resource issues, normal or not.
It is best to see the transfer time is very small when compared to the wait time; however, it is important to note that different access methods have different normal ratios of transfer to wait time. The ratio value itself is not importannt, but changes in this ratio, which are easy to see in this graph, may indicate some change or problem.
When running the X-Link Enterprise Edition, the graph does not display the message delivery times for all linkages, but rather the message delivery times for the linkage selected in the main dashboard window. This allows the administrator of an Enterprise Edition to manage each linkage individually.
So what does this graph mean?
The sample graphs above came from my development system and normal usage graphs would look a lot different. In a normal production system, we would expect to see a groups of message times based on hourly changes in volume based on the hours and/or service levels of the facility.
What we don't want to see is spikes or a consistent message deliver times that seem excessive. E.G., we're looking for outliers or simply crazy delivery times. Many things affect delivery times:
- The server is not responding or the connection to the server is not operational. This generally causes messages to be delayed for a very long time. It is generally not the fault of X-Link, but rather your service provider.
- The patient you are trying to update is locked by another user. This generally causes an additional 5 minute wait for the patient to become unlocked. If the patient remains locked for long periods of time, you will easily see these spikes on this graph.
- The delivery time of a local data base verses an internet API server will most likely be magnitudes greater. We're looking for unexplained spikes and consistent major changes, not at the specific delivery time itself.
What is important is to use the data from multiple graphs to get an overall picture of what is happening. This view may help you isolate where to look for issues.
Thank you. We hope this entry has helped you with the use of the Daily Message Counts Visualization.