Will Google Big Query Transform Big Data Analysis? by Doug Henschen.
From the post:
Google shared details Wednesday about Google Big Query, a cloud-based service that promises to bring the search giant’s immense compute power and expertise with algorithms to bear on large data sets. The service is still in limited beta preview, but it promises to speed analysis of Google ad data while opening up ways to mash up and analyze huge data sets from external sources.
Google Big Query was described by Ju-Kay Kwek, product manager for Google Cloud Platform Team, as offering an array of SQL and graphical-user-interface-driven SQL analyses of tens of terabytes of data per customer, yet it doesn’t require indexing or pre-caching. What’s more, customers will get fine-grained analysis of all their data without summaries or aggregations.
“Fine-grained data is the key to the service because we don’t know what questions customers are going to ask,” said Kwek in an onstage interview at this week’s GigaOm Structure Data conference in New York.
Some of Google’s beta customers are uploading data to the service with batches and data streams and treating it as a cloud-based data warehouse, but Kwek said ad data would be the first priority, supporting a Google customer’s need to understand massive global campaigns running in multiple languages.
“When an advertiser wants to understand the ROI or effectiveness of a keyword campaign running across the globe, that’s a big-data problem,” Kwek said. “They’re currently extracting data using the Adwords API, building sharded databases on-premises, doing all the indexing, and sometimes losing track of the questions they wanted to ask by the time they have the data available.”
Thus, time to insight will be the biggest benefit of the service, Kwek said, with analyses taking a day or less, rather than days or weeks, when customers face extracting and structuring data on less robust and capable on-premises platforms.
I am troubled by the presumptions that Google is making with Big Query.
Google’s Big Query presumes:
- Customer’s big data has value to be extracted.
- Value is not being extracted now due to lack of computing resources.
- The missing computing resources can be supplied by Big Query.
- The customer has the analysis resources to extract the value using Big Query. (Not the same thing as writing SQL or dashboards.)
- The customer can act upon the value extracted from its big data.
If any of those presumptions fail, then so does the value of using Google’s Big Query.
Resources for BigQuery developers. Including version 2 of the Developers Guide.