Search for: All records

Serverless applications represented as DAGs have been growing in popularity. For many of these applications, it would be useful to estimate the end-to-end (E2E) latency and to allocate resources to individual functions so as to meet probabilistic guarantees for the E2E latency. This goal has not been met till now due to three fundamental challenges. The first is the high variability and correlation in the execution time of individual functions, the second is the skew in execution times of the parallel invocations, and the third is the incidence of cold starts. In this paper, we introduce ORION to achieve these goals. We first analyze traces from a production FaaS infrastructure to identify three characteristics of serverless DAGs. We use these to motivate and design three features. The first is a performance model that accounts for runtime variabilities and dependencies among functions in a DAG. The second is a method for co-locating multiple parallel invocations within a single VM thus mitigating content-based skew among these invocations. The third is a method for pre-warming VMs for subsequent functions in a DAG with the right look-ahead time. We integrate these three innovations and evaluate ORION on AWS Lambda with three serverless DAG applications. Our evaluation shows that compared to three competing approaches, ORION achieves up to 90% lower P95 latency without increasing $ cost, or up to 53% lower $ cost without increasing tail latency. 

Serverless applications represented as DAGs have been growing in popularity. For many of these applications, it would be useful to estimate the end-to-end (E2E) latency and to allocate resources to individual functions so as to meet probabilistic guarantees for the E2E latency. This goal has not been met till now due to three fundamental challenges. The first is the high variability and correlation in the execution time of individual functions, the second is the skew in execution times of the parallel invocations, and the third is the incidence of cold starts. In this paper, we introduce ORION to achieve this goal. We first analyze traces from a production FaaS infrastructure to identify three characteristics of serverless DAGs. We use these to motivate and design three features. The first is a performance model that accounts for runtime variabilities and dependencies among functions in a DAG. The second is a method for co-locating multiple parallel invocations within a single VM thus mitigating content-based skew among these invocations. The third is a method for pre-warming VMs for subsequent functions in a DAG with the right look-ahead time. We integrate these three innovations and evaluate ORION on AWS Lambda with three serverless DAG applications. Our evaluation shows that compared to three competing approaches, \name achieves up to 90\% lower P95 latency without increasing \$ cost, or up to 53\% lower \$ cost without increasing P95 latency. 

We characterize production workloads of serverless DAGs at a major cloud provider. Our analysis highlights two major factors that limit performance: (a) lack of efficient communication methods between the serverless functions in the DAG, and (b) stragglers when a DAG stage invokes a set of parallel functions that must complete before starting the next DAG stage. To address these limitations, we propose WISEFUSE, an automated approach to generate an optimized execution plan for serverless DAGs for a user-specified latency objective or budget. We introduce three optimizations: (1) Fusion combines in-series functions together in a single VM to reduce the communication overhead between cascaded functions. (2) Bundling executes a group of parallel invocations of a function in one VM to improve resource sharing among the parallel workers to reduce skew. (3) Resource Allocation assigns the right VM size to each function or function bundle in the DAG to reduce the E2E latency and cost. We implement WISEFUSE to evaluate it experimentally using three popular serverless applications with different DAG structures, memory footprints, and intermediate data sizes. Compared to competing approaches and other alternatives, WISEFUSE shows significant improvements in E2E latency and cost. Specifically, for a machine learning pipeline, WISEFUSE achieves P95 latency that is 67% lower than Photons, 39% lower than Faastlane, and 90% lower than SONIC without increasing the cost. 

Interactive web services increasingly drive critical business workloads such as search, advertising, games, shopping, and finance. Whereas optimizing parallel programs and distributed server systems have historically focused on average latency and throughput, the primary metric for interactive applications is instead consistent responsiveness, i.e., minimizing the number of requests that miss a target latency. This paper is the first to show how to generalize work-stealing, which is traditionally used to minimize the makespan of a single parallel job, to optimize for a target latency in interactive services with multiple parallel requests. We design a new adaptive work stealing policy, called tail-control, that reduces the number of requests that miss a target latency. It uses instantaneous request progress, system load, and a target latency to choose when to parallelize requests with stealing, when to admit new requests, and when to limit parallelism of large requests. We implement this approach in the Intel Thread Building Block (TBB) library and evaluate it on real-world workloads and synthetic workloads. The tail-control policy substantially reduces the number of requests exceeding the desired target latency and delivers up to 58% relative improvement over various baseline policies. This generalization of work stealing for multiple requests effectively optimizes the number of requests that complete within a target latency, a key metric for interactive services.