Google updates App Engine datastore after outages

Google has introduced a new configuration of the database within its cloud platform Google App Engine, after periods of instability in its cloud infrastructure.

The High Replication Datastore (HRD) will provide heightened reliability and availability at the cost of write speed and changes in consistency guarantees within its API, Google announced on Wednesday through its Google Developer Blog.

"For the past six months, as you are probably aware, we've been struggling with some reliability issues with the App Engine Datastore. Over the course of the past few months, we've made major strides in fixing these issues," Kevin Gibbs of the Google App Engine Team wrote. "However, our experience with these issues has made us rethink some of our design assumptions."

Consequently, Google has now made available the HRD, which will provide increased availability for reads and writes to the datastore, but at the cost of increased latency for writes.

The HPD has increased availability by having more datacentres that hold replicas of the user's data, while employing a Paxos algorithm to sync data across all the locations in real-time. This means that application functionality will not be degraded during planned maintenance periods and "during most unplanned infrastructure issues", Google wrote.

Initial pricing is three times that of the default datastore until the end of 2011, after which Google expects pricing to change. Google recommends the HRD "primarily for developers building critical applications on App Engine who want the highest possible level of availability for their application".

Google App Engine is Google's answer to other cloud platforms offered by competitors such as Microsoft, Rackspace and Amazon Web Services (AWS).

Cloud architect JP Morgenthal told ZDNet UK he had experienced problems with applications he wrote for Google App Engine. "When I first deployed I saw strange messages coming to me about timeouts and performance implications, and I had to go back and modify my code to specifically handle the case where I was getting timeouts," he said.

Google's datastore "is nothing more than a basic MapReduce algorithm", Morgenthal said. "A Hadoop cluster on Rackspace is a better value proposition than on Google datastore.

"I don't particularly see where Google is really offering a tremendous service to attract customers when competing against the scalable cloud environments like Azure, Rackspace and AWS, where I can buy redundant storage and manage my [Hadoop] cluster with guaranteed rights of redundancy," he said.