With Cloud Bigtable, Google is launching a new NoSQL database offering today that, as the name implies, is powered by the company’s Bigtable data storage system, but with the added twist that it’s compatible with the Apache HBase API — which itself is based on Google’s Bigtable project. Bigtable powers the likes of Gmail, Google Search and Google Analytics, so this is definitely a battle-tested service
Google promises that Cloud Bigtable will offer single-digit millisecond latency and 2x the performance per dollar when compared to the likes of HBase and Cassandra. Because it supports the HBase API, Cloud Bigtable can be integrated with all the existing applications in the Hadoop ecosystem, but it also supports Google’s Cloud Dataflow.
Setting up a Cloud Bigtable cluster should only take a few seconds, and the storage automatically scales according to the user’s needs.
It’s worth noting that this is not Google’s first cloud-based NoSQL database product. With Cloud Datastore, Google already offers a high-availability NoSQL datastore for developers on its App Engine platform. That service, too, is based on Bigtable. Cory O’Conner, a Google Cloud Platform product manager, tells me Cloud Datastore focuses on read-heavy workload for web apps and mobile apps.
“Cloud Bigtable is much the opposite — is designed for larger companies and enterprises where extensive data processing is required, and where workloads are more complex,” O’Conner tells me. “For example, if an organization needs to stream data into, run analytics on and serve data out of a single database at scale – Cloud Bigtable is the right system. Many of our customers will start out on Cloud Datastore to build prototypes and get moving quickly, and then evolve towards services like Cloud Bigtable as they grow and their data processing needs become more complex.”
The new service is now available in beta, which means it’s open to all developers but doesn’t offer an SLA or technical support.
from jushiung1 http://feedproxy.google.com/~r/Techcrunch/~3/FrDTXR7tsdc/
via IFTTT
0 comentários:
Enviar um comentário