Whoscall Uses AWS Analytics to Improve Time-to-Market

By using AWS serverless architecture, Gogolook can develop new features 50–75 percent faster and save on development costs. Gogolook is the Taiwan-based company behind the international Whoscall caller identification app, which is supported by a database of 700 million phone numbers. The company uses Amazon Athena to run serverless queries, Amazon EMR to process raw data, and Amazon S3 as a data lake.

Many people have learned to recognize phishing emails or email scams, but phishing attempts via phone calls—known as vishing calls—can be harder to detect. The Whoscall app, developed by Taiwan-based Gogolook, instantly identifies callers through its database of more than 700 million phone numbers. Gogolook launched its app in 2013 with a mission to build a trusted global contact network for individuals as well as businesses.

To date, Whoscall has been downloaded to more than 80 million devices worldwide. Its main markets outside Taiwan are Japan, Hong Kong, and Thailand. Using its “collective phonebook” approach, Whoscall’s database is updated by user contributions as well as Gogolook’s partners, such as Google Maps. There is also a B2B aspect of the business, whereby companies can purchase personalized business cards that flash on mobile screens when they make calls.

Gogolook chose to build its app on the Amazon Web Services (AWS) Cloud thanks to the provider’s global presence and its local reputation for platform stability. Currently, Gogolook’s Whoscall app infrastructure runs from multiple AWS Availability Zones to ensure low application latency for users no matter where they reside. “We aim to provide information in 3 seconds or less under diversified hardware and network conditions,” explains Reiny Song, founder and chief technology officer at Gogolook. In the first half of 2020, the Whoscall app’s end-to-end network latency averaged less than 395 milliseconds, and uptime has consistently been at least 99.96 percent.

Analytics are an important part of improving the Whoscall service and attracting more app downloads. Engineers regularly review and update caller ID data, and Gogolook is always searching for ways to further automate analytics activities. Its Whoscall app collects user feedback and tracks app usage through session logs, storing 330 TB of raw data in an Amazon Simple Storage Service (Amazon S3) data lake.

The company uses Amazon EMR to process its big data, converting unstructured volumes into a format that data engineers can use for analysis. Previously, data engineers created an Amazon EMR cluster with an Apache Spark service to query the data directly from Amazon S3. This approach worked well, but data teams spent a lot of time monitoring and maintaining Spark clusters to improve efficiency. In the past year, however, Gogolook implemented Amazon Athena to quickly run interactive queries. With Amazon Athena, maintenance is now automated.

“The main benefit of using Amazon Athena is that our teams can use serverless queries to focus on data tests and querying the right data, instead of constantly performing system tests,” Song says. Data engineers can now pursue more complex, detailed queries and have 20 percent more time to design and perform new queries.

The shift to Amazon Athena was related to a larger move within Gogolook to integrate more serverless elements into its architecture. Part of Gogolook’s mission is to continually innovate using the latest technology to provide user-centric designs. In the past two years, Gogolook has introduced new features based on microservices, which are supported by serverless architecture. About 20 percent of its workloads are now serverless. According to Song, “Serverless is a great complement to our main workloads on traditional API interfaces.”

The new features rely on AWS Lambda to run code and Amazon API Gateway to create serverless APIs at scale. “If we want to set up a new service the traditional way, the process might take a month. Using Amazon API Gateway, we save at least half that time, so we can create something new in just one or two weeks,” Song says. Because of its quick build and takedown, as well as low hourly charging structure, developers are also encouraged to experiment more in the serverless environment. “The risk and cost are lower than the traditional way of developing, so our team members are not afraid to create and test new services.”

In addition to maintaining its database, Gogolook has set up another division of its data team tasked with creating new machine learning (ML) models. They are exploring ways to use ML to predict whether phone numbers are fraudulent or originate from telemarketing firms. The ML model will be able to quickly identify suspicious numbers so users can avoid answering such calls. The company is also considering how this service could benefit new B2B customers in financial services.

Gogolook consulted its AWS account team to discuss how its cloud architecture could best support new ML workloads. “We regularly evaluate which AWS services we can use to improve our data team’s performance,” Song says.

Recently, Gogolook upgraded from AWS Business Support to AWS Enterprise Support. This decision was prompted by the team’s positive experience with the comprehensive level of AWS support. For instance, during the implementation of Amazon Athena, Gogolook sought to customize the application limits to suit Whoscall workloads. Its dedicated AWS solutions architect recognized that Whoscall performs a high volume of queries in a short time frame and provided a direct line of communication with the Amazon Athena team to facilitate this. Song says, “AWS consultants spent a lot of time making sure they understood our situation and business use case and helped expedite communications and a resolution.”

The company has also benefited from an AWS Well-Architected Framework review to improve security and optimize spending. It estimates a savings of 20 percent on monthly infrastructure costs after implementing the recommendations that emerged from that review.

Since launching Whoscall six years ago, Gogolook’s employees have been focused on supporting data activities. Now that the business has grown its monthly active users from 100,000 to 10 million, owners recognized the need for a robust security review. “We benefit not only from AWS services to improve our product, but also from the reminders and assistance to set up strong security policies that protect our system,” Song says. The company relies on AWS Trusted Advisor to check access key rotation and Amazon GuardDuty to detect abnormal network behavior.

By 2021, Gogolook plans to establish its presence in at least one new market, grow its database, and build its business in Japan. “With our main AWS data center in the AWS Asia Pacific (Tokyo) Region, we expect to provide the best service to our Japanese users,” says Song. “AWS offers an infrastructure we can trust when we upgrade our production and scale Whoscall in the future.”

To learn more, visit aws.amazon.com/big-data/datalakes-and-analytics.