Audit your cloud infrastructure with Google Cloud Platform

MADGIC
HPC Industry & Services
Publié le 01 septembre 2022

Context

MADGIC is a mobile advertising revenue optimization platform that doubles the advertising revenues of mobile website publishers and mobile application developers.

To be effective, the MADGIC platform must have a very low processing time. A workaround is to limit latency by limiting computing power to 70%. ANEO was asked to audit the solution in order to remove this artificial and limiting limitation.

Challenge

The number of cores used to perform the calculation was too high, and therefore too costly. MADGIC processes around 300,000 offers per second, and this type of processing requires fast data access with very short response times and very low latency. To limit latency, computing power had to be limited to 70% of its capacity. Aneo had to find a solution to ensure optimum CPU utilization.

Solution

Using the Optimeo “Top-down” and “Bottom-up” method, ANEO's experts analyzed the different types of possible bottlenecks. In particular, they looked at the impact of core and hyperthreading configurations on GCP performance. They also studied the latency of Linux input/output sockets in order to reduce it as much as possible. They revisited the MADGIC GCP architecture, and probed the performance and use of databases via their programming interfaces.

Results

L’analyse d’ANEO a mis en lumière la présence de 5 goulots d’étranglement dans l’infrastructure d’application MADGIC. ANEO a conseillé de ne pas utiliser l’hyperthreading qui n’apporte pas de gain de performance. Une refonte de l’architecture de type « low reduce » a également été suggérée pour commencer le filtrage des annonces redondantes avant le traitement des annonces rentables par l’architecture MADGIC initiale.

ANEO a pu aider MADGIC à identifier des pistes d’optimisation de l’infrastructure et à conseiller des solutions à tester comme le filtrage des requêtes publicitaires avant le traitement pour réduire la pression sur l’application.

Ready to change how you work?

In the spotlight

cas clients - bioptimus
Accelerating LLMs on AWS Cloud for bioinformatics
06 June 2024 Artificial intelligence
Industry & Services
Aneo supported a bioinformatics company specializing in genAI/LLMs, in optimizing its AWS cloud infrastructure and AI processes to accelerate drug discovery.
Voir plus
cas clients - deeplife
HPC Workload Migration and Optimization to AWS
23 May 2024 AWS Migration
Industry & Services
DeepLife develops novel approaches to drug discovery based on state-of art multi-omics data, machine learning, and system engineering.
Voir plus
cas clients - okwin
Cloud migration to accelerate drug discovery with AWS
23 May 2024 AWS Migration
Industry & Services
Owkin is a leading biotechnology company specializing in the application of Artificial Intelligence (AI) and Machine Learning (ML) for drug discovery.
Voir plus