Google Launches Bigtable Hosting Service

Google yesterday announced the beta launch of Google Cloud Bigtable, a hosted version of its wide-column NoSQL database. The service is accessible via the standard HBase API–making it instantly integrated with the Hadoop stack–and blows away other hosted NoSQL offerings in terms of latency and throughput, the company claims.

You might know Bigtable as the database at the heart of many Google offerings, including Gmail, Google Search, YouTube, Google Earth, and Google Analytics. But the technology, which Google engineers Jeff Dean and Sanjay Ghemawat helped develop starting in 2004, also represents the first of a long line of distributed computing breakthroughs that includes Cassandra, HBase, Accumulo, Voldemort, and others.

“The systems they created and the white paper these guys wrote formed what I like to call the big bang of big data, both within Google and outside,” Google product manager Cory O’Connor said during a presentation at Strata + Hadoop World yesterday. “At Strata, many of the talks are based on systems that come directly from these white papers.”

Bigtable occupies a prominent position in the pantheon of big data tech

Now Google is offering the Bigtable database as a service, making it available to anybody with an Internet connection and an application that speaks HBase. For just $.65 per node per hour, developers can get access to a hosted Bigtable data store, and start storing terabytes or petabytes of data for their Web, mobile, and Internet of Things (IoT) applications.

Users will be amazed by Bigtable’s performance, O’Connor said. “Because it sits on Google’s core Bigtable technology, we’re able to realize some pretty impressive performance,” he said. “What you’re seeing is twice the price performance of any competing offering on the cloud right now in the same database class. And at the same time it gets single digit millisecond latency, 6 milliseconds latency, for both read and writes at the 99th percentile.”

Google sees its Bigtable Service being used in any industry that needs to store lots of data–including financial services, telecommunications, genomics, marketing and advertising technology, and anything the IoT touches. “Basically anywhere there are organizations dealing with very large data sets, moving them very quickly, trying to make them relevant, and trying to get these insights in real time,” O’Connor said.

Cloud Bigtable smokes the competition in read and write price/performance and latency

The Internet giant has several customers who are already using the product, including SunGard, which used Bigtable to build a financial audit trail system capable of ingesting 2.5 million trade messages per secon; Pythian, which integrated Bigtable with OpenTSDB provide a monitoring and metrics collection platform; CCRi, which integrated the open source spatio-temporal database “GeoMesa” with Bigtable to provide a scalable platform for real-time geospatial analysis in the cloud; and Telit Wireless Solutions, which integrated its IoT EAP (Application Enablement Platform) “m2mAIR” with Bigtable to boost data ingestion.

Cloud Bigtable is in beta, which is the same as general availability at regular software companies. To sign up, point your Web browser to cloud.google.com/bigtable.

Related Items:

Google Cloud Dataflow Now Open for Business

Google Bypasses HDFS with New Cloud Storage Option

Inside Sibyl, Google’s Massively Parallel Machine Learning Platform