It is supported with a more or less even, unidirectional force (the pull of the lift cells). However, simply hooking the tether right onto a straight keel would not produce an even load. Connecting the tether in the middle of the keel would put the most strain right at the point of contact. Distributing the tether's weight across the entire keel using cables connected to a load ring is a little better, but will put more tension in the longer cables. If we curved the keel, it could counteract this added tension with more compressive strength. Effectively, curving the keel increases the tension component parallel to the keel for the cables running to each end, at the expense of the perpendicular component.

The exact curvature will need to be determined. The typical method of doing this for a sailboat, airship or other bouyantly suspended vessel is to treat the entire hull as a single beam devided into discrete sections and then to determine the optimum curvature as one would for a truss supported only at those discrete sections. Fortunately for us, for the Virtual Beanstalk's keel this is not an approximation. The lift force of the balloons and downward tether tension is not distributed evenly across the keel struts but at discrete points, so there is little need for curvature of the struts themselves between load points. But analysis reveals that there are optimal angles for each of the straight struts within the keel. Even a less than "optimum" curvature (one that pays homage to actual design considerations of space and cross section rather than strictly following some force function)could improve the strength of the platform by reducing the amount it bends under a full load.