With the explosion of the size of digital dataset, the limiting factor for\ndecomposition algorithms is the \\emph{number of passes} over the input, as the\ninput is often stored out-of-core or even off-site. Moreover, we're only\ninterested in algorithms that operate in \\emph{constant memory} w.r.t. to the\ninput size, so that arbitrarily large input can be processed. In this paper, we\npresent a practical comparison of two such algorithms: a distributed method\nthat operates in a single pass over the input vs. a streamed two-pass\nstochastic algorithm. The experiments track the effect of distributed\ncomputing, oversampling and memory trade-offs on the accuracy and performance\nof the two algorithms. To ensure meaningful results, we choose the input to be\na real dataset, namely the whole of the English Wikipedia, in the application\nsettings of Latent Semantic Analysis.\n