AiThority Interview with Srini Srinivasan, CTO & Founder at Aerospike

Hi Srini, please tell us about your role and the team technology you handle at Aerospike.

As CTO of Aerospike, I’m challenged to ensure that our real-time data platform continues to meet the growing demand for real-time (microsecond) data decisioning at massive scale. Aerospike customer deployments have some of the most stringent SLAs across markets and use leading cloud and data infrastructure technologies to ensure we maintain five-nines availability for high-performance deployments at petabyte scale, which is of paramount importance to our customers and partners.

What kind of challenges did you / your team face during the pandemic? How did you leverage technology to overcome these challenges?

Like many companies, the initial loss of working physically together around a whiteboard proved to be a challenge for us and our customers. It was hard to work on projects and develop technology without being in the room together. Eventually, we were able to adapt by leveraging the remote collaboration technologies and developing the skills we are still using. Today, our teams have become more productive using a combination of in-person and virtual meetings effectively.

Recommended: Predictions Series 2022: AiThority Interview with Matt Watts, Chief Technology Evangelist, NetApp

Your take on the growing appetite for real-time data in the martech and adtech world – How does it align with the precision targeting and marketing every brand is looking to get hold of?

Real-time data is the lifeblood of digital advertising. Real-time bidding systems have millions of ad requests per second, and need low-latency access to lots of data to accurately respond to each one in milliseconds. Aerospike’s Real-time Data Platform allows superior use of third-party data and first-party data to better inform personalization and recommendation engines. This improves not only ad tech but other use cases, e.g., credit fraud, security breach detection, e-commerce, inter-bank money transfers, and so on.

How do you distinguish between historical data and real-time data to develop RTB systems for adtech platforms?

Aerospike can store complex document-oriented data records that include both real-time data and historical data. Additionally, there is a rich API for accessing both the real-time and historical data stored in these records. Aerospike applications can access these two types of data in a millisecond or less to make accurate decisions for real-time bidding under the 50ms required for an entire bidding cycle.

What kind of data ingestion/data integration technique do you use at Aerospike for solving real-time data challenges?

We use a patented hybrid memory architecture that leverages Flash storage to expand the real-time storage footprint per node. Using a log-structured file system, Aerospike supports extremely high throughputs for ingestion. For interoperability, Aerospike uses data connectors to enable integration with query and processing systems like Spark and Presto. Additionally, real-time connectors to Kafka, Pulsar, JMS, and Event Streaming make Aerospike an indispensable component in event streaming architectures.

Recommended: Gaming Predictions Series 2022: Welcome Metaverse. NFTs Will Get Hyped in Gaming

How are Big Data analytics coming together for the programmatic advertising ecosystem? How do these integrate with AI and machine learning capabilities?

Aerospike supports deployment at the core with vastly more data (petabytes) than at the edge (terabytes), and also supports asynchronous replication capabilities to move real-time data back and forth between the edge and the core.

The processing of the data in the core Aerospike database using Spark, for example, allows the generation of new AI/ML models based on the most recent activity on the edge. Furthermore, historical data and newly minted AI/ML models can move from the core to the edge, thus allowing the latest models to be applied at the edge to make accurate real-time decisions. These AI/ML use cases go far beyond ad tech.

Using the above, a continuous feedback loop can be implemented between model generation at the core followed by model application at the edge. The time taken for this feedback loop can be reduced significantly by using Aerospike both at the edge and core.

Tell us more about the ongoing race to compete for online ad bidding, placement, and delivery. Are we going to succeed with TV advertising tactics with mobile advertising breaking the adtech world into halves?

The online ad bidding race was started over a decade ago for display advertising, and Aerospike is recognized as a pioneer in that area. Today, the expansion in advertising and real-time bidding to channels like TV and mobile simply means that there are more opportunities for each of these channels to feed off each other.

For example, there will be new opportunities for identifying behavior by detecting multiple devices (i.e., users) in close proximity to a TV for an extended period (watching a show together, possibly). If you think of it as one screen across many devices, there is an opportunity for each device to gather data that will inform both TV and mobile advertising by sharing user behavior across multiple channels.

Recommended: AiThority Interview with Nelson Petracek, CTO at TIBCO Software

What are your predictions for the future of mobile adtech in 2022:

Adapting to the failures of ad cookies means more complex deployments. But the technology is there to process more data across more channels. Further, mobile ad tech will become more focused on privacy. Some predictions are:

Thank you, Srini! That was fun and we hope to see you back on AiThority.com soon

[To share your insights with us, please write to sghosh@martechseries.com]