For what it's worth, my opinion is that science needs peer review. There are some other options available, but they all seem to have problems of their own.
One option is for scientists to publish their work directly on the internet, on their own websites for example. However, you then have problems with locating the relevant information, and knowing whether it has any value (I came to this forum following someone who shares my name who published a complete load of nonsense as "science"). Very few scientists have the time to filter everything they read to spot errors. And also would the publishing scientist be diligent in putting links to rebutting articles on the site?
Another option is to have unreviewed journals. Even on the internet these would need some form of editorial input, which will result in papers being rejected by an editor who may have no particular knowledge of the subject.
Peer review allows scientists to take work done by other scientists and build on it or use it for other research without having to check every detail of that work themselves (naturally, there will be times when new research contradicts old papers, and so there will need to be critical reviews of both old and new work that needs to be done by a wider cross section of the scientific community than the usual peer reviewer or two).
What are the problems with peer review. There are three that I can think of off hand.
1) The "big name" problem mentioned by Joe Meert
2) The problem, especially in small fields, of all potential reviewers personnally knowing the author of the paper. Which can work two ways adversely affecting the process - the reviewer could be a mate or a rival of the author.
One thing that could work to counter these problems could be anonymous authorship; the paper is forewarded to the reviewers with the names of the authors removed. A paper will then stand on the quality of the science alone. It could be said that you can never entirely disguise the identity of the authors if they're known to the reviewer - but then again the authors can often make guesses about the identity of the reviewers at the moment anyway.
3) The "revolutionary idea" problem. Peer reviewers will have a tendancy to dismiss a radically new idea, it is only human nature. A radical idea will still need to be presented in a well written paper. And it is this class of paper that generally suffers from over exposure in the non-technical press.