This article is more than 1 year old

Cloudera whoops as its Hadoop loop-the-loops for cloud troupe

Get your Elephant-as-a-Service from IBM, Verizon, Savvis

Cloudera has teamed up with three midsize cloud operators to load its distribution of the Hadoop data muncher'n'cruncher into their clouds.

The alliances with IBM SoftLayer, Verizon Terremark, and Savvis CenturyLink were announced by Cloudera today, and will make the company's eponymous distribution of Apache Hadoop available as an installable image on the services.

"Historically a high fraction of our customers have been running [our software] on their own clusters in their own data centers because Hadoop tends to go where the data is. In the fullness of time more and more of those apps are running in the cloud," Cloudera's product veep Charles Zedlewski told El Reg. "We need to go to where that data is. More and more of those customers are asking for ways of running Cloudera in a cloud environment."

You can already spin up a Cloudera distribution on other clouds by loading it into your own virtual machines, he admitted, but via the partnership the vendors can directly resell it and cut the steps needed to get a Hadoop cluster up and running.

Prices for the service start at $699 per month on IBM's SoftLayer cloud, for example. In practice, the integration means Cloudera can now be provisioned directly via a web portal on the clouds and, typically, via APIs as well.

It could be possible for the clouds to offer deeper integration between their systems and Cloudera Hadoop by fiddling with low-level network management to better shuttle traffic between cloudy clusters, IBM confirmed, but it has no plans to do this at the moment.

The top-tier cloud providers – Amazon, Google, and Microsoft – are all absent from the scheme, probably because Amazon offers its own Hadoop variant (Elastic MapReduce), as does Microsoft via the HDInsight service and Google fields both stock Apache Hadoop along with MapR's high availability, distributed namenode version of the software.

"Customers are able to transfer their Cloudera subscription to Amazon today," he said. "That's a supported platform for us – it's just not a net new commercial relationship. In terms of a commercial model - our software drives a lot of consumption of amazon and we have production customers as well, we drive demand for EC2 instances and S3 storage."

He was mildly disparaging of the Amazon EMR and Microsoft's HDInsight services, claiming that though they offer a Hadoop capability they lack the secondary set of related services and technologies that Cloudera has with plugins like Impala.

"There is absolutely an audience for an EMR-like service," he said. "When I look at what Microsoft is doing in Azure, it feels a lot like an EMR knockoff... We're much more of an enterprise vendor."

There is some merit to this, as Cloudera's distribution comes with features not available in stock distributions due to the company's open core design approach.

However, cloudy elephant-logo'd Hadoop is likely a minority use case for the time being and is mostly of relevance to companies ingesting information through public-facing web apps. Furthermore, due to the bandwidth and time cost of pouring HDFS-stored data up into a cloud system, it's unlikely that either existing Amazon, Microsoft, or Google customers will be willing to migrate to another cloud.

Some on-premises enterprises may start slinging data up into one of these three cloud providers, and to our mind the best bet for this type of workload is Verizon Terremark due to the dedicated connectivity options the telco-cum-cloud company can serve up to elephant whispering punters. ®

More about

TIP US OFF

Send us news


Other stories you might like