More tech Stories
On The Web

Teradata’s CEO addressed the impact of Hadoop on its earnings call and, according to this report from ZDNet, downplayed its effect. In fact, he said only 4 to 8 percent of Teradata workloads might ever move to Hadoop. Even if that’s true for workloads, what about the data itself? It might not need to live in those pricey appliances.

In Brief

Dropbox has hired Kevin Park as its new head of technical operations and IT. Park was at Facebook from 2006 until 2011, where he was a director of technical operations. This isn’t the first time Dropbox has brought on former Facebook employees to help grow its engineering team — in 2012 it bought a startup called Cove that was started by Aditya Agarwal (now VP of engineering) and Ruchi Sanghvi (formerly VP of operations), who built Search and Newsfeed, respectively, during their time at Facebook.

Correction: This post has been updated to clarify that Ruchi Sanghvi is no longer with Dropbox. 

Upcoming Events

On The Web

This is a pretty interesting benchmark study, although the headline is a bit misleading because Hadoop isn’t really optimized for graph analysis. When you look at comparisons to Spark, GraphLab and other platforms, it seems the decision of what to choose might come down to data volume, acceptable latency and cost, especially when considered against the value of that graph workload. Projects like Giraph and other YARN-enabled engines might make Hadoop look better, too.

loading external resource
In Brief

I apologize if I’m late to the game on this, but someone just tweeted me about Apache Tajo, a potentially interesting new SQL query engine for Hadoop. I’m not sure how much traction it can possibly gain given the glut of other options out there (take a look at this now extremely outdated roundup from February), but I guess more options are better for users, to a point. SK Telecom, a Korean carrier, is already a big fan. Also, some of Tajo’s contributors’ employers are kind of interesting.

In Brief

Facebook has one of the largest, if not the largest, MySQL installations in the world, and has created a tool to keep that system online with as little human intervention as possible. It’s called MySQL Pool Scanner and, Facebook’s Shlomo Priymak wrote in a post on Monday describing it, it’s designed to automate “nearly everything a conventional MySQL Database Administrator (DBA) might do so that the cluster can almost run itself.” Not only does MPS handle availability but, Priymak noted, it also lets administrators do things such as copy the entire Facebook dataset with a single command.

In Brief

If you’ve ever wanted to use the Couchbase NoSQL database but didn’t feel like managing servers, a San Mateo, Calif.-based startup called KuroBase says it has you covered with its new service. Cloud databases are already pretty popular with web developers running MongoDB, Postgres and even CouchDB (kind of, technically), but I believe this is a first for Couchbase. It could be popular, though, especially if developers are keen on Couchbase’s new ability to sync data between mobile devices and a central database.

In Brief

We have been hearing about things like YARN and high availability for a few years — they’ve even been incorporated into some commercial Hadoop distributions — but now they’re finally part of the official Apache Hadoop code base. Technically version 2.2.0, “The project’s latest release marks a major milestone more than four years in the making, and has achieved the level of stability and enterprise-readiness to earn the General Availability designation,” according to an Apache Software Foundation press release.

On The Web

I think this is more about Hadoop and other emerging technologies than the analysts quoted here are willing to admit. Why do you think Teradata is pushing its Hadoop story so much lately? There is, for example, crazy excitement around big data and Hadoop in China. Customers with blank slates center their efforts around Hadoop, while big existing customers are trying to offload more to Hadoop. Teradata sales are fairly flat right now even in the U.S. because big existing customers are getting bigger but fewer are signing up.

On The Web

IBM has shared some details about a new project called WatsonPaths that lets doctors actually interact with the system to understand how it came to its conclusions, and to tell it whether its “thinking” was right. This type of interaction is critical in any type of machine learning system where speed isn’t the primary objective, because it lets humans see things they might not have and also train the machines to be more accurate in the future. WatsonPaths is a GUI-based tool and is being developed along with doctors at the Cleveland Clinic.

15678944page 7 of 44