I’m not so sure the problem is the slider controls so much as the labeling and self-documentation of your application classes, plus perhaps the fact that you are using percents.
You have Essential and Non-Essential application classes. Doesn’t that cover pretty much everything? I mean, if something is not an Essential, then isn’t that by definition non-essential? Okay, apparently there is something called “Business Relevant”, which is somehow between Essential and Non-Essential, and then there’s Other, which is the class the users give too few percent to.
So what’s Other? Whatever it is, it’s not essential, and not business relevant, and is, somehow, even less important than non-essential. So what is it? Superfluous applications? Unused services? Bloat? Spam? Malware? “Other” sounds like that category of miscellaneous stuff in a pie graph that no one really cares about and tends to account for only a few percent anyway. So maybe the problem is labeling a category as “Other” suggests it should only be a few percent or even zero.
So consider relabeling your application classes:
Make the natural order of the classes more apparent. Frankly, simply calling them (only) Priority 1, 2, and 3 may be better than having any qualitative labels at all like Essential, Business Relevant, and Non-Essential (unless these are existing terms of art for your users).
If you really think qualitative labels will help users choose values, consider something that is non-exhaustive like Essential, Critical, and Important (leaving, implicitly, unimportant for Other).
Better, or additionally, include text that succinctly explains what applications are in each class. This could be a sample of specific applications (e.g., “Zoom to Head Branch”), or a brief definition of what constitutes each class (e.g., “Priority 1: Applications that, if interrupted, would force the business to shut down completely”).
Most important, re-name the Other category, making it clear with its name and description why it’s important and why it should in fact have most of the bandwidth. For example “Normal: Applications used in day-to-day operations of the business (majority of typical bandwidth)”; or “Dynamic: Bandwidth automatically allocated to Essential, Critical, Important, and anything else as needed.”
Using percents may be a problem because it’s not the units the user really thinks of or uses for minimal bandwidth. Users also know that percents go from 0 to 100, so that may subtly suggest they need a middling number like 30%, not something “extreme” like 8%.
Your use case implies that users know the minimal bandwidth in Mbps of each application. If so, use that instead, rather than making users do arithmetic to sum all Mbps from a class than divide by the total bandwidth. Perhaps for each application the user declares as Priority 1 (and Priority 2), the user also specifies its minimum guaranteed bandwidth in Mbps. Let the system sum it to calculate the minimal guarantee for the class (assuming that’s even needed)
The one problem I can see with the design with slider controls is that it’s hard to see how the value of one class affects the values of other classes (assuming that the classes and their minimal bandwidth are mutually exclusive and additive). The four-color bar at the bottom helps, but it would be better to integrate that graphic in the controls themselves. You may want to replace all three sliders with a single multi-range slider, where there’s a single slide with three handles, each dividing the bandwidth among four application classes (including Other). That may make it clearer. It would also give equal attentional weight to Other because its value becomes directly manipulated, rather than a footnote on the page.
However, sliders are best when users are entering values to just a “looks good” criterion. If users arrive at the UI with specific values in mind (e.g., 28.6% or 22 Mbps), rather than just, “oh, about that much,” then it’s actually easier for the user to type the values into some text boxes that make them try to drag a slider to an exact value.
If you do go with text boxes, consider putting the Other class (with its improved labeling) prominently at the top, and the three editable text boxes for the other classes below so users can see they're "stealing" from Other when they set their minimums.