Google has taken a big leap into the Infrastructure-as-a-Service (IaaS) game, announcing Google Compute Engine on the second day of its Google I/O conference in San Francisco.
“This goes beyond just giving you greater flexibility and control,” Craig McLuckie, product manager for Google Compute Engine, wrote in a June 28 corporate blog posting. “Access to computing resources at this scale can fundamentally change the way you think about tackling a problem.”
Internally, Google routinely tackles the biggest of Big Data issues—indexing the Internet on top of serving billions of search queries per day requires massive data centers. Google Compute Engine gives regular users access to that processing power, allowing them to launch clusters of enormous size.
Specifically, users can launch Linux virtual machines on-demand, with 1,2, 4, and 8 virtual-core virtual machines available. Each virtual core offers 3.75GB RAM. Users can either store their data locally, or rely on Google Cloud Storage, which will cost a set amount per GB:
Other Google-derived storage options include ephemeral disk, “a simple block device” linked to the lifecycle of the virtual machine; when the VM stops, the data from this type of disk is lost.
A second option is persistent disk, a network-connected storage service, where data is replicated across multiple physical disks in a Google data center. “You can also create snapshots of your disks for backup/restore purposes, and can mount these devices in a mode that allows multiple virtual machines to read from a single device,” reads Google’s explanation. “As with our ephemeral disk, all data is encrypted at rest.”
Users can connect their virtual machines into a network in order to create clusters, control them via scriptable command-line tools or Web UIs (all tools are built on RESTful API, with plans to release tools to open-source), and configure firewalls for security. Google is apparently working with RightScale, Puppet Labs, and other providers to deliver familiar tools to the ecosystem. Google Compute Engine is currently in limited preview.
This isn’t Google’s first foray into giving businesses the tools for wrestling with giant datasets. In May, the company rolled out a limited preview of BigQuery, a service that lets users apply a “SQL-like query language” to analyze up to billions of data rows in seconds. BigQuery features secure SSL access, group- and user-based permissions via Google accounts, and the capability to scale to trillions of records.