You have some options: gel filtration in Sephadex G-25 (fast and efficient), a mixed-bed resin or dialysis. If you choose desalting using Sephadex, GE and other companies (Pierce, Thermo Scientific, etc.) offer several columns ready to use.
One option is to use ultrafiltration devices you can buy from several suppliers like Milipore(Amicon) or Pall. There are several cutoffs available from 5 up to 300 kDa, so you have to choose a cutoff smaller than your protein(s) of interest. With these devices you can exchange the buffer by centrifugation, get a high recovery and can even concentrate the sample. We use them as a standard for buffer exchange and/or concentration.
As Mirco says, those devices can work fine. However, It depends on the protein; you can run a test. We had some trouble with ultrafiltration by centrifugation with some proteins; in those cases the yield at the end drops significantly, even if the membrane doesn't leak. Clive Dennison (A guide to protein isolation) has an equation that allows to calculate how much will decrease salt concentration after "n" washings: Xn = Xo [u/(u+v)]^n where Xn is salt concentration after "n" washings, Xo is the original salt concentration, u is the final volume after centrifugation, and v is the added volume of new solution.
I agree with Gustavo. Recently I had an cytochrome c solution and tried to get rid of a low molecular weight reagent it was incubated with (the buffer was just water. All ended up with a red colored membrane and very poor recovery. I guess it was due to the high pI of cytochrome c (9,6) and the polyethersulfone-membrane. So some kind of cation exchange chromatography rather than ultracentrifugation. The use of 1M ammonium acetate pH 3 has helped to recover most of the cytochrome c.
The gel filtration or dialysis may not suitable for your case, as you mentioned that it is a restriction enzyme. I assume that you purchased the restriction enzyme and not expressed and purified by yourself. As Miraco said, buffer exchange by centrifugal filters can work fine.