Data Sheet—Monday, August 11, 2014

August 11, 2014, 1:18 PM UTC

Welcome to the Monday edition of Data Sheet. Would you pay $200,000 for someone with just two years experience? That’s the starting salary some companies are willing to pay for the “high priests of algorithms,” a.k.a. data scientists. Plus: you know all those impossible-to-remember passwords you make your employees pick? They may not be any harder to break than simple one.


A hint of GE's big data master plan. Last year, it plunked down $105 million for a stake in Pivotal. Now, it's using the software for predictive aircraft maintenance and to streamline operations management for other GE industrial customers.

John McAfee rants: Beware of smartphones. In the strange-but-true column, the once-respected antivirus pioneer came out of seclusion to explain why he no longer uses his eponymous software and why he's content with "stupid phones." 

More complex passwords aren't necessarily more secure. All those bizarre alpha-numeric combinations are supposed to make for unbreakable credentials. But do they really? Not so much. Oh, and about those 1.2 billion stolen passwords we got excited over last week? Respected security expert Bruce Schneier has joined the experts debating whether the numbers were inflated.

Informatica's inflection point. Just two deals in the data integration company's latest quarter came from client in what is usually its biggest vertical, financial services. Its longtime CEO riffs on healthcare innovation and why it wants to be the "Switzerland of all things data." Plus, more healthcare tech talk from insurance giant Humana's CEO.

NetScout's not happy with Gartner "pay to play" research model. Its network performance technology didn't get a top rating on a recent "magic quadrant," so it's suing for defamation. A similar suit five years ago was settled. Which makes me ask: Do CIOs really buy technology based on reports like these?


Hey! You! Get out of my cloud. In late July, a U.S. judge ruled emails stored in Microsoft's Ireland data center were subject to a U.S. search warrant. Google, Microsoft and Twitter are rallying behind Facebook in a similar New York privacy dispute. Between hackers, feds and marketers, everyone wants their hands on your cloud data. IBM isn't leaving things to chance. It just got a patent that lets business more closely manage where cloud data lives geographically, so you avoid onerous regulations.


Big bang for Chromebooks. Sales could triple to 14.4 million by 2017 from 2.5 million this year. Samsung's the top dog, but Hewlett-Packard is winning businesses in banking, real estates and hospitality.

Facebook's balancing act squeezes power consumption. Engineers devised an automated approach for shuffling server production loads in the social network's data center that cuts its energy bill 10% to 15% over a 24-hour period. Here's how it works.

Will future Mac OS versions take diction? A 92-page patent that hints at the talkative computer in the film Her details interfaces for a desktop digital assistant akin to the iPhone's Siri. Just don't expect it to sound like actress Scarlett Johansson. 


Special delivery for IronSource. Most of the details are secret, but the Israeli software company, which enables digital downloads, just got another $85 million ahead of an anticipate second quarter 2015 IPO.


Your network has plenty of important things to tell your IT operations team, that is, if you're willing to listen. So far, analytics startup ExtraHop has raised in $61.1 million in venture capital for technology to help write that story.

ExtraHop is pioneering real-time analytics for wire data—basically all the L2-L7 packets and transactions streaming back and forth between computers to run applications. It sells both a hardware appliance that you can stick in your data center plus a cloud-delivered service. The technology starts around $10,000 annually for the cloud version.

Sounds like your basic network probe, right? So I checked in with cofounder and CEO Jesse Rothstein to ask what makes ExtraHop's platform unique.

Here's his short answer: "We're pushing the envelope of how much processing you can do in real time, and still keep up with these conversations. Traditional network probes just count bits and packets and answer questions like, 'Is there communication? What words are used?' To provide the level of analysis that we do, where we're actually understanding what the conversation actually means, is probably five or six evolutions beyond that." 

Why does this matter? Consider the experience of ExtraHop customer Concur, which deals with something like 2 billion SQL queries daily. The company was considering a seven-figure network upgrade because of performance issues until ExtraHop discovered the problem was from a bad script bogging down bandwidth.

ExtraHop doesn't supersede technologies like Splunk, which analyze machine data, it complements it. In fact, ExtraHop is adding two new services this week to its Open Data Stream portfolio: connectors with the MongoDB and Elasticsearch analytics engines. It already integrates with Splunk and VMware's log insights dashboards.

Rothstein notes: "One of our customers said it best when he said, 'Wire data and machine data are like chocolate and peanut butter.' One of the reasons they go so well together is that machine data is inherently based on self-reporting. You only know about problems if the system logs that there is a problem. You only know that something is slow if an application logs that something is slow. This is very valuable, but in an enterprise IT environment, there's all the stuff you can log, and then there's everything else. … Wire data is observational. We know about a problem because we observe them occurring on the wire, we observe the failures."

The question is, are you listening?


All hail the data scientist! Companies like LinkedIn, Facebook, TaskRabbit and Intuit are paying $200,000 to $300,000 a year for entry-level candidates. No wonder so many mathematics and physicists are turning their backs on academia. Still trying to figure out what to prioritize when it comes to your company's data projects? Don't worry, you're not alone, but here are five priorities for whomever you appoint to be chief data officer: leverage, enrichment, monetization, protection and upkeep.


Gartner Catalyst: Architect a digital business. (Aug. 11 – 14, San Diego)

VMworld: Learn about latest virtualization innovation. (Aug. 24-28, San Francisco)

Atlassian Summit: Build software, collaboratively. (Sept. 9 – 11, San Jose, Calif.)

Open Data Center Alliance Forecast 2014: Catch up on enterprise cloud trends. (Sept. 22 – 24, San Francisco)

Oracle OpenWorld: Get a roadmap reality check. (Sept. 27 – Oct. 2, San Francisco)

Gartner Symposium ITxpo 2014: Compare notes. (Oct. 5 – 9, Orlando, Fla.)

Splunk .conf2014. Glean intelligence from machine data. (Oct. 6 – 9, Las Vegas)

Dreamforce: Pick from 1,400 sessions about the world's largest cloud ecosystem. (Oct. 13-16, San Francisco)

Strata/Hadoop World: Analyze big data tools and techniques. (Oct. 15 – 17, New York)

AWS re:Invent: Hear the latest about Amazon Web Services. (Nov. 11 – 14, Las Vegas)

Gartner Data Center Conference: Get new ideas for operations and management. (Dec. 2 – 5, Las Vegas)