In the past, online ad networks owned their supply of impressions for ads with publishers and matched it to their own demand (advertisers). However, in the past few years, with programmatic ad selling becoming common place, these ad networks are now appealing to demand from multitude of demand side platforms (DSPs). The general approach is to partition the demand into pools (based on type of ads or the demand party). An inevitable challenge is the impedance mismatch between the marketplace and the demand pools. Thus, a central problem is to figure out how to handle each supply request and selectively send requests to demand pool so as to optimize the performance of entire supply. This is the Callout Problem. We present a large scale data analysis and control system that (a) continually learns changing traffic patterns, capacities, bids, revenue etc, and (b) picks the best slice of traffic to send to each demand pool, subject to their capacity constraints. The optimization is based on a greedy solution to the underlying knapsack problem which easily adapts as capacities change over time. We have implemented and deployed this solution for the callout problem in one of the largest mobile ad marketplaces(InMobi) and has been operational for several months. In this paper, we will describe the scale of the problem, our solution and our observations from operational experience with it. We believe a well-engineered solution to the Callout problem is essential for many ad networks in the online ad world.

In this paper, we present DLSH Distributed Locality Sensitive Hashing , a similar-data search technology. The huge growth in the size of video content has broken the traditional multi-media index hosting and look-up solutions, these are not able to scale to the size of the current and projected index requirements. Distributed LSH (D-LSH) addresses this need of a highly scalable multi-media index. DLSH performs better for finding approximate near neighbors on extremely large scales , as DLSH distributes close points on single boxes, and far points on different boxes based on projections.

vADeo is a next-generation video ad system that analyzes the video to find appropriate scene changes where ads can be inserted. The context of each ad insertion point is determined through high-level analysis of the surrounding video segment thereby making ads contextual. Further vADeo implements two novelties on the player side - ad book-marking and delayed interaction - which encourage ad clicks without disrupting the video viewing experience.