After stepping on a rake in a privacy fiasco last week, Evernote is looking ahead to finishing its cloud migration, and revving up machine learning and natural language processing to improve the service for users.
With the days ticking down on the calendar, Evernote expects to hit its self-imposed deadline and complete migration by year's end from its own data centers to Google Cloud, Evernote Corp. CTO Anirban Kundu tells Light Reading.
Evernote announced its plans to migrate in October, citing desires to cut costs, speed upgrades and improve scalability. (See Why the Evernote Elephant Packed Its Trunk for Google Cloud.)
The migration takes three parts: Incoming network requests need to be redirected to the Google Cloud Platform; about 4.5 petabytes of user content, including notes and attachments, need to be relocated; and Evernote needed to make some software changes to get to the cloud, Kundu says.
Software changes include transitioning data form blob stores to Google Cloud Storage, and making use of upgraded versions of Pub/Sub and other services available from Google, he says.
Evernote has two data centers, which are leased. While it retains a small portion of operations in those data centers, 99% of the service is moving to Google Cloud, Kundu says.
The migration required Evernote to learn new methods of user management, separating production versus development and access control, he says.
Evernote anticipates that new machine learning technologies, enabled by Google Cloud, will help make the service more useful to users.
For example, many people store travel itineraries in Evernote, and the company hopes machine learning will be able to identify notes related to travel itineraries, and surface that information to the user in the most convenient format, time and location. Today, a user viewing notes in a snippet view sees a text excerpt of the note; instead, the user might see a short summary of a travel itinerary. When at an airport, the mobile Evernote app might show notes related to the itinerary first, rather than seeing notes in the default chronological order. And Evernote might automatically inform the user of a delayed flight, Kundu says.
Also, many people store meeting notes in Evernote; the service might automatically identify to-dos and action items, determine who those actions were assigned to and ping the user how the action should be handled, Kundu says.
Not on the table: Using machine learning to sell products or serve up ads. "We are looking at this exercise from a 100% user benefit," Kundu said. The company generates all its revenue from user subscriptions, both from individuals and from enterprise licenses.
Evernote's reliance on subscriptions -- and its attempts to break into the enterprise market -- make last week's privacy snafus particularly problematic.
Following the initial furor, Evernote posted not one, not two, but three separate updates walking the announcement back.
Evernote at first clarified that notes viewed by employees would be anonymized to protect users. Then it decided to cancel the privacy change entirely, saying it would no longer allow employees to read notes except when legally required do to so, such as in cases where Evernote receives a subpoena or court order or needs to prevent or mitigate violations to terms of service -- standard for cloud providers.
According to a blog post by CEO Chris O'Neill on Monday, Evernote plans to upgrade its internal privacy controls and processes. "While I'm confident that our existing controls provide a level of security comparable with other cloud providers, it has become clear to me that parity is not enough," O'Neill says. "Our goal should be to lead the thinking in this area, as new technologies continually challenge society's understanding and expectations of what privacy is and should be."
- Evernote CEO: 'We Messed Up' on Privacy
- Evernote: 'We Communicated Poorly' on Privacy Change
- Evernote Caught in Privacy Imbroglio
— Mitch Wagner, , Editor, Light Reading Enterprise Cloud