Qubole rains down analytic insights from Hadoop cloud

Facebook-style tools for business development

Elastic fantastic

These customers have enabled Qubole to run a B-round and raise another $13m this year, taking total funding to $20m. At a press briefing in Mountain View in December Qubole said QDS had processed 500 petabytes of data for its users, with an 86PB/month current rate; a year ago it was 7PB/month, and it could reach 100PB/month this year.

It’s spinning up 10 - 1,500-node clusters per day and 250,000 virtual machines a month. The largest cluster so far totalled is 1,800 nodes.

Pinterest Qubole query and user growth

“[Qubole] competes with Amazon EMR, Hortonworks and Cloudera for big data analysis in the cloud [with] no need for local data centre," said Gil Allouche, the company's marketing veep.

Allouche also thinks it makes more sense to run Cloudera on-premise as it’s too rigid to run in the cloud.

“[Customers] write queries and Qubole spins up Hadoop cluster in the cloud. Initially on AWS but expanded onto Google cloud platform - it’s the main Hadoop platform there - also onto Azure.”

It agreed with Microsoft to make QDS available to Azure customers in December last year.

Allouche said Qubole is:

• Compliant with OpenStack but there’s no demand for it. We see more and more enterprise customers coming from Azure.
• On Azure marketplace soon. For now, you go to Qubole website and the SW is deployed on cloud of choice. We handle cloud admin behind the scenes.
• We're not using HDFS in standard form. We save data in S3 in high tables using Open Source HIVE.
• Data ingest is via Qubole tools ... ODBC and JDBC connectors and also ingesting files for CSV files and data stored in S3, Azure buckets, and Google.
• The vast majority of data we work on is in a log store ... most if it ends up in S3. Connectors allow you to pull data from other sources and join it into one set.
• We have an SDK and an API fully available to developers.

“We help customers plan their big data spending for a month. Query run time can be capped to prevent cost over-run [and] Query can be billed to a business department," said Allouche.

Elastic pricing starts at $500/month.

If that’s too expensive customers can move into a subscription with a committed number of hours of compute - you only pay for compute. Customers already pay for storage on cloud already. They only use Qubole for analytics.

One target application area for big data analytics is fraud detection in online banking, retail, healthcare, telecommunications and online advertising - detecting bot-driven clicks for example.

Qubole has about 70 staff overall, with 15 - 20 in Mountain View and others in Banglalore, such as the CTO and much engineering. It’s hiring 2-3 people a week and looks to have a smart and well-timed HaaS offering with money to develop its product and build out its marketing. Watch this space.

Oh, and the name; where did that come from?

It may be relevant to know that Q’bole is slang for “what’s up?” in Spanish. ®