Grid Bandwidth Considerations

Infoblox Grid technology relies upon database replication for its core functionality. When designing a Grid, it is important to consider the amount of traffic generated by this replication and the overall number of Grid members. Other communication between Grid members (such as log retrieval and monitoring functions) occurs as well. All of this traffic is securely communicated between the Grid Master and Grid memb ers through encrypted VPN tunnels.
One component of the traffic through the tunnels is database replication traffic. There are three types to consider:

• Complete database replication to a Master Candidate — Occurs when a Master Candidate joins or rejoins a Grid. The Grid Master sends the complete database to a Master Candidate so that it has all the data it needs if it ever becomes promoted from member to master.
• Partial database replication — Occurs when an appliance or HA pair joins or rejoins the Grid as a regular member (which is not configured as a Master Candidate). The Grid Master sends it the section of the database that mainly applies just to the member.
• Ongoing database updates — Occurs as changes are made to the Grid configuration and data. The Grid Master sends all ongoing database updates to Master Candidates and individual member-specific updates to regular members.

If there are no or very few DNS dynamic updates, and no or very few DHCP lease offers and renewals issued, then this type of replication traffic is minimal.
If there are many DDNS (dynamic DNS) updates (many per second) and/or many DHCP lease offers and renewals (many per second), then the replication traffic is the largest component of the VPN traffic among Grid members.

Note: A Grid Master replicates data to single members and to the active node of HA members. The active node then replicates the data to the passive node in the HA pair.

At a minimum, there must be 256 Kbps (kilobits per second) bandwidth between the Grid Master and each member, with a maximum round-trip delay of 500 milliseconds. For ongoing database updates, the amount of data sent or received is 15 Kb for every DDNS update, and 10 Kb for every DHCP lease -offer/renew. The baseline amount for heartbeat and other maintenance traffic for each member is 2 Kbps. Measure the peak DNS and DHCP traffic you see in your network to determine the bandwidth needed between the Grid Master and its members for this activity.
For example, you might decide to place your Grid members in the locations shown in Figure 5.6.

Figure 5.8 Grid Deployment

In this example, the Grid Master is optimally placed in the Data Center West. There are a total of seven members: the HA Grid Master, three HA members, and three single members. If all the members are Master Candidates, the Grid Master replicates all changes to the other six members. Assuming that the master receives 20 dynamic updates per minute and 40 DHCP lease renews per minute, the calculation for Grid bandwidth is:

Another component is the upgrade process. See Upgrading NIOS Software for more information.
Bandwidth requirements, database size, and update rate determine the maximum size of the Grid you can deploy. Based on the various factors discussed above, you can determine the amount of bandwidth your Grid needs. If your calculations exceed the available bandwidth, then you might need to modify your deployment strategy, perhaps by splitting one large Grid into two or more smaller ones.

Note: This calculation does not take into account existing traffic other than DNS and DHCP services, so factor and adjust accordingly.

For international networks, because of bandwidth and delay requirements, a geographical grouping of Grid members might be the best approach. For example, if you have a global presence, it may make the most sense to have a North American Grid, a South American Grid, a European Grid, and an Asia/Pacific Grid.