RuffCT is an upgrade to the existing RingCT MLSAG ring signatures. RuffCT is an informal name in honor of Tim Ruffing, who is one of the authors of the paper (soon to be published) from which this new type of ring signature originates.
Ring signatures are what give Monero untraceability. Therefore ring signatures that allow for higher ring sizes will mean greater untraceability, privacy and fungibility for Monero. Ring signatures that are more compact will give Monero greater scalability. RuffCT delivers these improvements.
RingCT ring signatures are O(n) with respect to storage requirements, where n is the ring size. RuffCT ring signatures are O(log n) in size. In our tests, this means a ring signature with 1024 inputs will only take twice as much space as a ring signature with 8 inputs, instead of requiring 128x the storage.
RuffCT ring signature sizes are also independent of the number of real inputs. A RingCT ring signature with 20 real inputs and a ring size of 32 (meaning 32 * 20 = 640 inputs in total referenced in the ring signature) would require 41600 bytes of storage in RingCT. An equivalent RuffCT ring signature would only require 802 bytes of storage. As you can see, these are dramatic storage savings.
These savings are storage savings, and not computation savings. It has therefore been suggested that Monero use GPU accelerated Elliptic Curve math to allow for higher ring sizes. GPU acceleration can mean up to 120x faster EC operations with modern consumer graphics cards, which would pave the way for higher ring sizes that are more computationally intensive.
 GPU accelerated EC operations paper: https://eprint.iacr.org/2014/198.pdf