Type I' string theory is equivalent to M-theory compactified on a line segment times a circle, i.e. M-theory on a cylinder.
M-theory on a line segment only is the Hořava-Witten M-theory, a dual description of the $E_8\times E_8$ heterotic string, because every 9+1-dimensional boundary in M-theory has to carry the $E_8$ gauge supermultiplet. The extra compactified circle is needed to break the $E_8\times E_8$ gauge group to a smaller one; and to get the right number of large spacetime dimensions, among other things.
Type I' string theory has D8-branes that come from the end-of-the-world branes in M-theory on spaces with boundaries; it also possesses orientifold O8-planes. Interestingly enough, the relative position of O8-planes and D8-branes in type I' string theory may be adjusted. This freedom goes away in the M-theory limit; the D8-branes have to be stuck at the orientifold planes, those that become the end-of-the-world domain walls of M-theory, and this obligation is explained by the observation that an O8-plane with a wrong number of D8-branes on it is a source of the dilaton that runs. In the M-theory limit, the running of the dilaton becomes arbitrarily fast which sends the maximum tolerable distance between the O8-plane and D8-branes to zero.
This post imported from StackExchange Physics at 2014-03-09 09:13 (UCT), posted by SE-user Luboš Motl