New Relic Insights allows developers to harvest real-time statistics about running apps and crunch the results in its cloud
Big data provides insight into items with many moving parts, and few objects have more moving parts than the applications used in any given enterprise. Today New Relic, a creator of enterprise application monitoring tools, is unveiling a real-time data analysis service for applications that harvests data directly from inside all those moving parts.
New Relic Insights, as it's now called (it wasformerly known as Rubicon), could be described as "Splunk for apps," but Lew Cirne, CEO of New Relic, calls it "software analytics."
Insights consists of two parts: a set of analytics extensions added to the application and a cloud-based service -- built on top of a multitenant database system -- that ingests the harvested data from the apps and allows detailed querying, analysis, and reporting. Applications written in Ruby, PHP, Java, .Net, and Python are all supported, and support for Node.js is coming soon, according to Cirne.
To examine the harvested data, the developer interacts with Insights via NRQL, a SQL-like query language. A query like
SELECT count(*) FROM PageView since 12 weeks ago would return the total number for 12 weeks' worth of page views for a given app. Other keywords, like FACET(NRQL's version of the SQL GROUP BY keyword), can be used to further slice and dice data.
Some might consider Insights a long-term replacement for Splunk, but Cirne doesn't see it that way. Splunk is a complementary technology focused on analyzing output from log files, as opposed to harvesting data directly from inside apps, he said. "At the end of the day, Splunk helps IT people answer questions for sysadmins. If an app happens to log data in the right way, Splunk might be useful there. Our customers are developers, and our vantage point is the software itself."
A key area where New Relic has broken away from many of its big-data analytics brethren is how Insights was put together on the back end. The entire service is a custom, closed source creation and not a repurposing of an existing open source data solution like Hadoop.
Cirne believes that off-the-shelf database solutions weren't built from the inside out for the kind of transactional speed and real-time data ingestion and processing that New Relic wanted, nor were they designed to scale in the kind of back-end multitenant architecture his company wanted to create for its customers.
"Products like Hadoop cover many more use cases than timestamped key/value events, and have different accuracy requirements," Cirne said. "In our case, if I say there's 3.67 thousand users, if I'm off by 10 that's not going to matter for analytics. But you can't build a generic database that way because it might be balancing bank accounts.
"By building Insights for these kinds of queries, and for multitenants, we've built something that delivers high speed. And that isn't possible, we think, on any open source technology we've looked at," Cirne said.
No comments:
Post a Comment