How are you reading the points from your strategy? Every time you read a point in your strategy, the control engine does a network read which will include all the TCP/IP overhead. How you write your strategy determines the “scan rate” - how often the strategy requests the IO.
To reduce the amount of bandwidth, you need to read your IO less (increasing the loop delays, perhaps). Another way to reduce bandwidth, while also increasing performance, (at the cost of complexity) you can look into using the MoveIoUnitToNumTable or MoveIoUnitToNumTableEx commands. These commands will perform a block read from the brain, significantly decreasing the network overhead. Do this one time in your chart, like right after your charts delay. You also are in control of how often you read your variables through the loop delays in your chart (You do have delays, right?).
Are you running an HMI, PAC Display? If you are, you can set it to read the IO from a local variable in your strategy (like the table from the MoveIoUnitToNumTable command) so the HMI isn’t making any requests to the remote IO.