There was someone(it is not know, if it was the same person) trying that some weeks ago and he failed pretty hard. Then he tried it again and failed again and now we have this from maybe the same person or a copycat.
So, there is a lot of preparation to that and you need some servers. The whole thing is also not cheap, when it comes to transaction fees.
Also keep in mind, that some people are already working on counter-measures.
The first "stress test" was organized informally by a crowd call on /r/bitcoin. Each volunteer spammed what and when it felt like, during a specified time window.
The second "stress test" was organized by a mysterious entity called "coinwallet.eu". They said they had ~5000 euros to spend on fees. that attempt failed after spending ~500 euros because their servers broke under the load. The testalso wasn't effective because they were using low fees.
This test is their new attempt, with more servers and smarter scripts, paying higer and adaptive fees. They said that their aim was to get 250 MB of unconfirmed transactions on the queues; which, with 1 MB blocks filled 50%, will take ~10 days to clear after they stop the test.
They are already at 150 MB, still climbing. Yesterday they managed to drive up the "effective minimum fee" (EMF) that you would need to pay for prompt service, to ~100'000 satoshis/kB, or ~0.4 mBTC (~0.11 USD) for a typical transaction. (Maybe more, the plots that I am watching broke down at that point.) Now they seem to be using lower fees, because the EMF is less than half of that even though the backlog is much bigger.