Cluster Range Distribution Policy

A probe cluster is a logical container for a number of Data Flow Probes. When you define a network range for the cluster, the cluster distributes IP ranges to its Probes, thereby ensuring that the load on the Probes is evenly balanced.

A cluster balances its IPs among its Probes as follows:

  1. To ensure that specific IPs are always distributed to a particular Probe, you can set a TQL query constraint on that Probe. When the TQL query runs, the resulting IPs are distributed to the respective Probe.

  2. All of the remaining IP addresses in the cluster's range are then distributed evenly among all of the cluster's Probes (including the Probes with the TQL query constraints).

    Note The cluster balances its IPs over its Probes in such a way so as to minimize IP movement after the IPs have been distributed to the Probes.

    If the cluster's IPs become unbalanced, the cluster tries to re-balance the IPs among its Probes. The cluster is allowed to retry to balance the Probes depending on how much IP movement is allowed. For details, see Limit IP Movement in a Cluster.

  

Example

Consider the following scenario:

  • Cluster Range: 10.10.10.0 - 10.10.10.255

  • Cluster includes Probe1, Probe2, Probe3

  • TQL constraint on Probe1: Probe1's range must include IPs 10.10.10.20 - 10.10.10.40

After running a TQL query on RTSM, the following IPs are returned:

The cluster balances its network ranges among its Probes as follows:

  • Because of the TQL constraint on Probe1, range 10.10.10.30 - 10.10.10.40 is specifically distributed to Probe1
  • Range 10.10.11.0 - 10.10.11.20 is out of the cluster's range, so it is not distributed to any of the cluster's Probes
  • The remaining IPs in the cluster's ranges are then evenly balanced among all of the cluster's Probes. The remaining ranges are:

    • 10.10.10.0 - 10.10.10.29
    • 10.10.10.41 - 10.10.10.255

    The cluster always attempts to distribute an IP address to a Probe whose TQL constraint includes that IP address

    Note Technically, IPs 10.10.10.20 - 10.10.10.29, which fall within Probe1's TQL constraint, can be distributed to any of the cluster's Probes, and not necessarily only to Probe 1. This is because these addresses were not found by the TQL query, which means they do not exist in RTSM. In turn, no actual workload can be distributed to them anyway, so it is insignificant which Probe received these IP addresses.

    Nonetheless, the cluster does make a special effort to distribute these IPs to Probe1. This way, if a subsequent recalculation of the TQL query returns these IPs, they are already in place, and IP movement is minimized.

    Furthermore, when the TQL query is recalculated, and if these addresses are found in RTSM, these IP addresses are indeed distributed to Probe 1, per its TQL constraint.