Understanding resource-balancing behavior

When you set the resource-balancing behavior, you balance the amount of moves made during resource balancing with the amount of accuracy achieved. Accuracy is how successfully the moves were made, based on the number of moves allowed. The higher the accuracy, the more evenly resources are balanced.

You also specify the location of the Domino® Change Control database (DOMCHANGE.NSF). By default, Activity Trends automatically selects a server. However, you must specify the Domino Change manager server in the Configuration Settings document. Use the default unless you want to use a local replica or are working remotely and want to use a server that has a replica of the Domino Change Control database.

Resource balancing distributes database activity across three bins:

  • Light -- The first bin, when graphed, has the lightest amount of activity.
  • Medium -- The second bin, when graphed, has a medium amount of activity. This percentage is calculated based on the percentage in the other two bins.
  • Heavy -- The third bin, when graphed, has the heaviest amount of activity.

Resource balancing attempts to balance the bins among the servers as well as the total for the servers. This is important because heavily utilized databases (databases with a high number of transactions) also have the greatest variance. That is, their usage is more likely to vary from the mean more frequently. This means that when there is a spike in activity, the spike will be a big spike, and the dip will be a big dip. Dividing the databases into bins separates the few databases that account for a large amount of activity, from the large amount of databases that account for little activity. For example, out of 100 databases on a server, 10 databases may account for 30% of activity, while 65 databases account for another 30%. The remaining 40% of activity is accounted for by the medium usage 25 databases.

Balancing according to the bins, ensures that the spread of heavily used and lightly used databases are evenly distributed across the servers. This results in more predictable usage patterns, increased availability, and more efficient use of resources.

Deciding the exact percentages for each of the bins depends on how your organization uses its databases and the type of server being balanced (mail server versus application server). For mail servers in most organizations you may want to increase the size of the light bin and decrease the size of your heavy bin, while for application servers the mix may be different.

You can view profile charts on bin activity and weigh percentages to improve load balancing.

You also specify how Activity Trends analyzes the server resource capacities. By default, server capacities are determined relative to other servers in the list. For example a server that has a capacity of x1 transactions has half the transactional capability (CPU) of a server at x2. You could, however, balance resources based on actual values (such as the number of transactions per day, or the total amount of disk space available). Using the previous example, you would specify the servers as having a capacity of 10,000 and 20,000 transactions. However, if you choose to balance resources based on actual values, you have to know that the servers involved can actually handle the capacities specified.

Another way in which you indicate server resource capabilities, is to specify how the server volume is determined. You can either use server volume and file system information when resource balancing, or ignore volume information and treat all space as flat. The default is to use the volume information, which uses the different physical volumes and their sizes that comprise the space available to Domino, rather than just the total amount of space on the server. Volume balancing is recommended. This may produce plans in which a database moves to a different server and has a different destination path because of space requirements on a particular volume on the destination server.