Using Intel® hardware, Intel® Integrated Performance Primitives (Intel® IPP) and Intel® oneAPI Data Analytics Library (oneDAL), Anodot improved the performance of its autocorrelation function (ACF) and XGBoost algorithms, significantly reducing machine learning (ML) compute time and costs associated with autonomous business monitoring and anomaly detection.
The data analytics company created a solution for its customers that identifies revenue-critical business incidents in real time through models that analyze hundreds of millions of time series metrics every minute. For the anomaly-detection platform, unlimited scalability and effective management of compute costs are needed as it grows – in addition to improving upon the speed, efficiency and accuracy of model training and inferencing.
While Anodot already runs its AI platform on Intel® CPUs, the team ran performance tests on the Intel® Xeon® Scalable processor platform in an extended collaboration. Through optimizations to ACF using Intel IPP for anomaly detection, the team recorded up to 127 times faster training performance and a 66% reduction in the overall cost of running the training algorithm in a cloud environment – achieved by cutting the ACF runtime by almost 99%. Optimizations to XGBoost algorithms using oneDAL and the baseline XGBoost model for forecasting resulted in 4 times faster inferencing time, as well as enabling the service to analyze 4 times the amount of data at no additional cost for inference.1
“When choosing a machine learning platform, you need to think about scale as your business grows,” said Ira Cohen, chief data scientist at Anodot. “So, model efficiencies and compute cost effectiveness become increasingly important. Our performance tests show the Intel software and Xeon platform provide us efficiency gains that will allow us to deliver an even higher quality of service at lower cost.”
More: Read the full case study, “Anodot Optimizes Anomaly Detection Services.”
1 By cutting the ACF runtime by almost 99%, the runtime of the overall algorithms is reduced by almost 66%. Assumptions provided by Anodot. Instance cost/CPU hour (4 vCPU) $0.050; # CPU hours per 1 M metric 500; # Training runs per month 10. Reference white paper for detailed information regarding tests conducted by Anodot.
Anodot ran performance tests on AWS instances in November 2021. More detailed information can be found in this Intel white paper: “Accelerate Real-Time Machine Learning Based Anomaly Detection and Forecasting at Scale”
Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy.