Amazon teaches cloud to speak Pig Latin
Adoophay orfay ethay assesmay
Amazon has taught its cloud to speak Pig Latin.
In April, Jeff Bezos and company unveiled a new web service based on Hadoop, the open-source phenomenon that seeks to mimic MapReduce, the distributed data-crunching platform that drives Google's online infrastructure. And today, Amazon announced that its Elastic MapReduce service now includes support for Pig Latin, the Hadoop programming language first developed at Yahoo!.
You can write straight to Hadoop in Java, but Pig puts Hadoop programming on a somewhat higher level. As Amazon puts it: "Pig Latin is a SQL-like data transformation language. You can use Pig Latin to run complex processes on large-scale compute clusters without having to spend time learning the MapReduce paradigm."
But the SQL comparison is a tad misleading. Hive - a Hadoop programming language seeded by Facebook - is closer to SQL, as is a second as-yet-unnamed language under development at Yahoo!. Pig sits somewhere between the SQL-like paradigm and the low-level code of MapReduce.
Amazon offers two means of using Pig: an "Interactive mode," which lets you run Pig queries on an existing MapReduce cluster by setting up an secure shell connection, and a "batch mode," which involves launching multiple MapReduce server instances that reference your Pig Latin.
Based on Google-published research papers, Hadoop mimics the company's MapReduce framework, which maps data-crunching tasks across distributed machines, splitting them into sub-tasks, before reducing the results into one master calculation. Thus Amazon Elastic MapReduce.
The Apache-hosted Hadoop was originally developed by Nutch-crawler founder Doug Cutting. After three and half years at Yahoo! developing the platform, Cutting is now headed for Cloudera, a Silicon Valley startup that has commercialized Hadoop - Red Hat-style. ®