Twitters data cannot fit on one machine. In 2015 their Hadoop cluster was 30 PB per earlier comments/their blog. How do you fit that on one machine?