{"id":21627,"date":"2018-12-06T15:40:07","date_gmt":"2018-12-06T23:40:07","guid":{"rendered":"https:\/\/insidebigdata.com\/?p=21627"},"modified":"2018-12-06T15:41:13","modified_gmt":"2018-12-06T23:41:13","slug":"better-faster-graph-processing","status":"publish","type":"post","link":"https:\/\/insidebigdata.com\/2018\/12\/06\/better-faster-graph-processing\/","title":{"rendered":"Better, Faster Graph Processing"},"content":{"rendered":"

MIT CSAIL’s new programming language GraphIt outperforms other models 5x while reducing lines of code<\/em><\/p>\n

Every time you Google sometime, you\u2019re benefiting from something called graph data. Google represents everything on the web as a giant graph of different webpages and their relationships to each other. (This is in contrast to representing data as essentially a series of rows and columns.)<\/p>\n

So-called \u201cgraph processing\u201d is used for all sorts of things in our lives, from Uber driving directions to Facebook \u201cPeople You May Know\u201d friend recommendations. However, there have been many limits in terms of being able to use graph processing on a larger scale. As a result, programmers often have to use different combinations of techniques that make trade-offs in terms of efficiency and flexibility.<\/p>\n

However, a team from MIT CSAIL has developed a new programming language for graph processing that could help. Dubbed \u201cGraphIt<\/a>,\u201d the new domain-specific language has been shown to outperform existing frameworks by a factor of nearly 5x while also reducing the lines of code by almost an entire order of magnitude.<\/p>\n