Just wanted to give some updates on the stuff Chronos just mentioned. Trigonometric parallax has, so far, only given us distances to objects within ~100 parsecs (~300 light years). Beyond that, the parallax angle becomes too small for our instruments to resolve. However, the upcoming Space Interferometry Mission (SIM) is going to be able to do microarcsecond astrometry, which means that it could, in principle, measure the distance to anything in the galaxy (within its limiting magnitude, of course).
Beyond the distances that can be measured by parallax, we must use "secondary" methods. This includes the standard candles that Chronos mentioned. The reason they should be thought of as secondary (or, in some cases, tertiary or higher) is that they must be calibrated by some other distance-finding method. That is, we don't know the intrinsic brightness of a standard candle unless we can measure the distance and flux to one nearby. This means, unfortunately, that higher-order distance-finding methods carry with them the systematic errors of the lower-order ones.
The standard candle that can take us furthest (so far) is the Type Ia supernova. In principle, this can be used to measure the Hubble constant and normalize the distance-redshift relationship that is the subject of this thread. However, it turns out that Cepheids are actually better for this job. Why? Well, the basic reason is that we can get better statistics with Cepheids -- there aren't enough supernovae occurring nearby. However, supernovae are much brighter than Cepheids, so they can take us to much larger distance and are much better for measuring the higher-order changes in the distance-redshift relationship. This is why we were able to detect the acceleration of the universe with them.