Lazar Cvetković
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Lazar Cvetković has a total of 18 co-authors. Here are the top co-authors by number of shared publications:
- Ana Klimovic
ETH Zurich, Zurich, Switzerland
5 shared publications - Mihajlo Djokic
ETH Zurich, Zurich, Switzerland
2 shared publications - Andrea Lattuada
MPI-SWS, Saarbrücken, Germany
1 shared publication - Boris Goranov
ETH Zurich, Zurich, Zurich, SWITZERLAND
1 shared publication - Boris Grot
University of Edinburgh, Edinburgh, United Kingdom
1 shared publication - Dmitrii Ustiugov
NTU Singapore, Singapore, Singapore
1 shared publication
Latest Publications
- Open Access
- 19 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 860-866
Towards a Lightweight Sidecar-based Service Mesh for Serverless
- 19 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 860-866
Open AccessTowards a Lightweight Sidecar-based Service Mesh for Serverless
- Open Access
- 12 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 944-961
Unlocking True Elasticity for the Cloud-Native Era with Dandelion
- 12 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 944-961
Open AccessUnlocking True Elasticity for the Cloud-Native Era with Dandelion
- Open Access
- 4 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 369-384
Dirigent: Lightweight Serverless Orchestration
While Function as a Service (FaaS) platforms can initialize function sandboxes on worker nodes in 10-100s of milliseconds, the latency to schedule functions in real FaaS clusters can be orders of magnitude higher. The current approach of building FaaS cluster managers on top of legacy orchestration systems (e.g., Kubernetes) leads to high scheduling delays when clusters experience high sandbox churn, which is common for FaaS. Generic cluster managers use many hierarchical abstractions and internal components to manage and reconcile cluster state with frequent persistent updates. This becomes a bottleneck for FaaS since the cluster state frequently changes as sandboxes are created on the critical path of requests. Based on our root cause analysis of performance issues in existing FaaS cluster managers, we propose Dirigent, a clean-slate system architecture for FaaS orchestration with three key principles. First, Dirigent optimizes internal cluster manager abstractions to simplify state management. Second, it eliminates persistent state updates on the critical path of function invocations, leveraging the fact that FaaS abstracts sandbox locations from users to relax exact state reconstruction guarantees. Finally, Dirigent runs monolithic control and data planes to minimize internal communication overheads and maximize throughput. We compare Dirigent to state-of-the-art FaaS platforms and show that Dirigent reduces 99th percentile per-function scheduling latency for a production workload by 2.79× compared to AWS Lambda. Dirigent can spin up 2500 sandboxes per second at low latency, which is 1250× more than Knative.
- 4 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 369-384
Open AccessDirigent: Lightweight Serverless Orchestration
While Function as a Service (FaaS) platforms can initialize function sandboxes on worker nodes in 10-100s of milliseconds, the latency to schedule functions in real FaaS clusters can be orders of magnitude higher. The current approach of building FaaS cluster managers on top of legacy orchestration systems (e.g., Kubernetes) leads to high scheduling delays when clusters experience high sandbox churn, which is common for FaaS. Generic cluster managers use many hierarchical abstractions and internal components to manage and reconcile cluster state with frequent persistent updates. This becomes a bottleneck for FaaS since the cluster state frequently changes as sandboxes are created on the critical path of requests. Based on our root cause analysis of performance issues in existing FaaS cluster managers, we propose Dirigent, a clean-slate system architecture for FaaS orchestration with three key principles. First, Dirigent optimizes internal cluster manager abstractions to simplify state management. Second, it eliminates persistent state updates on the critical path of function invocations, leveraging the fact that FaaS abstracts sandbox locations from users to relax exact state reconstruction guarantees. Finally, Dirigent runs monolithic control and data planes to minimize internal communication overheads and maximize throughput. We compare Dirigent to state-of-the-art FaaS platforms and show that Dirigent reduces 99th percentile per-function scheduling latency for a production workload by 2.79× compared to AWS Lambda. Dirigent can spin up 2500 sandboxes per second at low latency, which is 1250× more than Knative.
- Open Access
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Enabling In-Vitro Serverless Systems Research
Serverless is an increasingly popular cloud computing paradigm that has stimulated new systems research opportunities. However, developing and evaluating serverless systems in a research setting (i.e., "in-vitro", without access to a large-scale production cluster and real workloads) is challenging yet vital for innovation. Recently, several serverless providers have released production traces consisting of large sets of functions with their invocation inter-arrival time, execution time, and memory footprint distributions. However, executing the workload synthesized from these traces requires a massive cluster, making experiments expensive and time-consuming. In this work, we show how to use the data available in production traces to construct workload summaries of configurable scales that remain highly representative of the original trace characteristics and can be used to evaluate serverless systems in-vitro. Compared to random sampling of functions from the original trace, our method can generate summaries of up to 10X higher representativity, measured as the average of the Wasserstein distances of the distributions of interest (e.g., function execution time and invocation inter-arrival time) from the respective distributions in the original trace. We release our toolchain that enables researchers to synthesize representative workload summaries and show how it can be used to evaluate the performance of serverless systems at diverse load scale factors.
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessEnabling In-Vitro Serverless Systems Research
Serverless is an increasingly popular cloud computing paradigm that has stimulated new systems research opportunities. However, developing and evaluating serverless systems in a research setting (i.e., "in-vitro", without access to a large-scale production cluster and real workloads) is challenging yet vital for innovation. Recently, several serverless providers have released production traces consisting of large sets of functions with their invocation inter-arrival time, execution time, and memory footprint distributions. However, executing the workload synthesized from these traces requires a massive cluster, making experiments expensive and time-consuming. In this work, we show how to use the data available in production traces to construct workload summaries of configurable scales that remain highly representative of the original trace characteristics and can be used to evaluate serverless systems in-vitro. Compared to random sampling of functions from the original trace, our method can generate summaries of up to 10X higher representativity, measured as the average of the Wasserstein distances of the distributions of interest (e.g., function execution time and invocation inter-arrival time) from the respective distributions in the original trace. We release our toolchain that enables researchers to synthesize representative workload summaries and show how it can be used to evaluate the performance of serverless systems at diverse load scale factors.
- Open Access
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 22-28
Understanding the Neglected Cost of Serverless Cluster Management
Serverless computing enables the cloud platform to optimize resource management under the hood to improve performance and resource-efficiency. However, today's serverless cluster managers are designed by simply retrofitting legacy workload orchestration systems, despite the unique characteristics of serverless workloads. We study Knative-on-K8s as a representative state-of-the-art cluster manager for serverless and show that it can cause second-scale delays and contribute to over 65% of end-to-end latency for function invocations experiencing cold starts. These overheads occur when the cluster experiences high sandbox churn, which is common in production serverless deployments. We analyze the root cause of current cluster manager overheads for serverless workloads and propose a set of design principles to improve end-to-end latency and peak throughput by rethinking the cluster manager system architecture.
- 23 October 2023
- Published by Association for Computing Machinery (ACM)
- p. 22-28
Open AccessUnderstanding the Neglected Cost of Serverless Cluster Management
Serverless computing enables the cloud platform to optimize resource management under the hood to improve performance and resource-efficiency. However, today's serverless cluster managers are designed by simply retrofitting legacy workload orchestration systems, despite the unique characteristics of serverless workloads. We study Knative-on-K8s as a representative state-of-the-art cluster manager for serverless and show that it can cause second-scale delays and contribute to over 65% of end-to-end latency for function invocations experiencing cold starts. These overheads occur when the cluster experiences high sandbox churn, which is common in production serverless deployments. We analyze the root cause of current cluster manager overheads for serverless workloads and propose a set of design principles to improve end-to-end latency and peak throughput by rethinking the cluster manager system architecture.