Christian Pinto
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Christian Pinto has a total of 27 co-authors. Here are the top co-authors by number of shared publications:
- Archit Patke
University of Illinois Urbana-Champaign, Urbana, USA
2 shared publications - Haoran Qiu
University of Illinois at Urbana-Champaign, Urbana, USA
2 shared publications - Ravishankar K. Iyer
University of Illinois at Urbana-Champaign, Urbana, Illinois, USA
2 shared publications - Saurabh Jha
IBM Research, Yorktown Heights, New York, USA
2 shared publications - Zbigniew Kalbarczyk
University of Illinois at Urbana-Champaign, Illinois, USA
2 shared publications - Angelos Bilas
Institute of Computer Science, Foundation for Research and Technology - Hellas (FORTH), Heraklion, Crete, Greece and Computer Science Department, University of Crete, Heraklion, Crete, Greece
1 shared publication
Latest Publications
- Open Access
- 27 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 246-253
GeoServe: Leveraging Disaggregated Data Processing for Scalable Geospatial Model Serving
Geospatial foundation models (GFMs) operate on large, multi-band raster products (e.g., GeoTIFF) that require expensive data access and preprocessing - reprojection, decoding, normalization, and tiling - before GPU inference. In our measurements, reading and preprocessing geospatial inputs can be orders of magnitude slower than tokenization or standard image preprocessing, and constitute 31 - 43% of end-to-end request time for a representative GFM. Existing inference frameworks such as vLLM execute this preprocessing inline with request handling, which under load serializes CPU and I/O work, increasing queueing delay, and leaving GPUs underutilized. We present GeoServe, a Ray-based serving system that decouples the geospatial data pipeline from GPU inference by disaggregating I/O- and CPU-heavy preprocessing to a scalable pool of CPU workers, while keeping GPU nodes dedicated to model forward passes. We show experimentally that GeoServe reduces the p90 request latency by up to 414.9× at high load and improves throughput by up to 4.74× compared to vanilla vLLM, while increasing the achieved model forward-pass rate from ~ 16 inf./sec to ~ 72 inf./sec via better batching opportunities.
- 27 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 246-253
Open AccessGeoServe: Leveraging Disaggregated Data Processing for Scalable Geospatial Model Serving
Geospatial foundation models (GFMs) operate on large, multi-band raster products (e.g., GeoTIFF) that require expensive data access and preprocessing - reprojection, decoding, normalization, and tiling - before GPU inference. In our measurements, reading and preprocessing geospatial inputs can be orders of magnitude slower than tokenization or standard image preprocessing, and constitute 31 - 43% of end-to-end request time for a representative GFM. Existing inference frameworks such as vLLM execute this preprocessing inline with request handling, which under load serializes CPU and I/O work, increasing queueing delay, and leaving GPUs underutilized. We present GeoServe, a Ray-based serving system that decouples the geospatial data pipeline from GPU inference by disaggregating I/O- and CPU-heavy preprocessing to a scalable pool of CPU workers, while keeping GPU nodes dedicated to model forward passes. We show experimentally that GeoServe reduces the p90 request latency by up to 414.9× at high load and improves throughput by up to 4.74× compared to vanilla vLLM, while increasing the achieved model forward-pass rate from ~ 16 inf./sec to ~ 72 inf./sec via better batching opportunities.
- Open Access
- 27 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 13-22
Sampling Where It Matters: Predicting LLM Serving Performance
- 27 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 13-22
Open AccessSampling Where It Matters: Predicting LLM Serving Performance
- Open Access
- 8 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1206-1218
Page Migration for Hardware Memory Disaggregation Across a Network
- 8 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 1206-1218
Open AccessPage Migration for Hardware Memory Disaggregation Across a Network
- Open Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 23-33
An I/O Characterizing Study of Offloading LLM Models and KV Caches to NVMe SSD
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 23-33
Open AccessAn I/O Characterizing Study of Offloading LLM Models and KV Caches to NVMe SSD
- Open Access
- 20 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 18-35
Queue Management for SLO-Oriented Large Language Model Serving
Large language model (LLM) serving is becoming an increasingly critical workload for cloud providers. Existing LLM serving systems focus on interactive requests, such as chatbots and coding assistants, with tight latency SLO requirements. However, when such systems execute batch requests that have relaxed SLOs along with interactive requests, it leads to poor multiplexing and inefficient resource utilization. To address these challenges, we propose QLM, a queue management system for LLM serving. QLM maintains batch and interactive requests across different models and SLOs in a request queue. Optimal ordering of the request queue is critical to maintain SLOs while ensuring high resource utilization. To generate this optimal ordering, QLM uses a Request Waiting Time (RWT) Estimator that estimates the waiting times for requests in the request queue. These estimates are used by a global scheduler to orchestrate LLM Serving Operations (LSOs) such as request pulling, request eviction, load balancing, and model swapping. Evaluation on heterogeneous GPU devices and models with real-world LLM serving dataset shows that QLM improves SLO attainment by 40-90% and throughput by 20-400% while maintaining or improving device utilization compared to other state-of-the-art LLM serving systems. QLM's evaluation is based on the production requirements of a cloud provider. QLM is publicly available at https://www.github.com/QLM-project/QLM.
- 20 November 2024
- Published by Association for Computing Machinery (ACM)
- p. 18-35
Open AccessQueue Management for SLO-Oriented Large Language Model Serving
Large language model (LLM) serving is becoming an increasingly critical workload for cloud providers. Existing LLM serving systems focus on interactive requests, such as chatbots and coding assistants, with tight latency SLO requirements. However, when such systems execute batch requests that have relaxed SLOs along with interactive requests, it leads to poor multiplexing and inefficient resource utilization. To address these challenges, we propose QLM, a queue management system for LLM serving. QLM maintains batch and interactive requests across different models and SLOs in a request queue. Optimal ordering of the request queue is critical to maintain SLOs while ensuring high resource utilization. To generate this optimal ordering, QLM uses a Request Waiting Time (RWT) Estimator that estimates the waiting times for requests in the request queue. These estimates are used by a global scheduler to orchestrate LLM Serving Operations (LSOs) such as request pulling, request eviction, load balancing, and model swapping. Evaluation on heterogeneous GPU devices and models with real-world LLM serving dataset shows that QLM improves SLO attainment by 40-90% and throughput by 20-400% while maintaining or improving device utilization compared to other state-of-the-art LLM serving systems. QLM's evaluation is based on the production requirements of a cloud provider. QLM is publicly available at https://www.github.com/QLM-project/QLM.
- Open Access
- 12 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 1507-1511
Sunfish: An Open Centralized Composable HPC Management Framework
Traditional HPC systems are provisioned with static fixed quantities of memory, storage, accelerators, and CPU resources to execute requested computation. This is not sufficient for today’s datacenters that are running modern dynamic workloads, resulting in workloads executing on systems that are not optimized for their needs. Datacenters often end up over-provisioning hardware resources to attempt enabling any workload to run on any server in the cluster. Composable Disaggregated Infrastructure (CDI) enables servers to be composed out of physically disaggregated resources to support the requirements of a given workload. Resource management, using a standardized interface, can enable clients to monitor, compose, and intelligently provision resources, in beneficial ways. The OpenFabrics Alliance in collaboration with DMTF, SNIA, and the CXL Consortium, is developing the Sunfish Management Framework. The goal of Sunfish is to enable interoperability through common interfaces for connecting workloads with resources, without having to worry about the underlying network technology.
- 12 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 1507-1511
Open AccessSunfish: An Open Centralized Composable HPC Management Framework
Traditional HPC systems are provisioned with static fixed quantities of memory, storage, accelerators, and CPU resources to execute requested computation. This is not sufficient for today’s datacenters that are running modern dynamic workloads, resulting in workloads executing on systems that are not optimized for their needs. Datacenters often end up over-provisioning hardware resources to attempt enabling any workload to run on any server in the cluster. Composable Disaggregated Infrastructure (CDI) enables servers to be composed out of physically disaggregated resources to support the requirements of a given workload. Resource management, using a standardized interface, can enable clients to monitor, compose, and intelligently provision resources, in beneficial ways. The OpenFabrics Alliance in collaboration with DMTF, SNIA, and the CXL Consortium, is developing the Sunfish Management Framework. The goal of Sunfish is to enable interoperability through common interfaces for connecting workloads with resources, without having to worry about the underlying network technology.
- 4 January 2022
- Published by Taylor & Francis
- p. 165-190
Data Convergence for High-Performance Cloud
The distributed computing landscape has been undergoing radical changes: high-performance computing (HPC) applications are moving to the cloud, as a way of simplifying development, deployment, and migration across computing systems. Meanwhile, cloud applications are becoming increasingly complex and computationally intensive, with the advent of high-performance data analytics pipelines – highly distributed workflows dealing with enormous and diverse datasets, heavily relying on virtualization and containers, which would benefit from technologies used in “traditional” HPC. The foreseeable convergence demands new abstractions to cope with the increased heterogeneity, so that diverse workload classes can coexist seamlessly on the same infrastructure. In this chapter, we propose a unified storage layer, to enable cloud-native applications to transparently access a wide spectrum of storage solutions, ranging from high-performance filesystems to cloud-based object stores and key–value databases. Allowing software to exploit the best out of the storage services available with no need for manual intervention from users or programmers, simplifies development and execution of workflows, and boosts the overall productivity. This work has been motivated by the requirements of real-world, industry-driven applications running on Kubernetes, the industry standard for cloud infrastructure orchestration.
- 4 January 2022
- Published by Taylor & Francis
- p. 165-190
Data Convergence for High-Performance Cloud
The distributed computing landscape has been undergoing radical changes: high-performance computing (HPC) applications are moving to the cloud, as a way of simplifying development, deployment, and migration across computing systems. Meanwhile, cloud applications are becoming increasingly complex and computationally intensive, with the advent of high-performance data analytics pipelines – highly distributed workflows dealing with enormous and diverse datasets, heavily relying on virtualization and containers, which would benefit from technologies used in “traditional” HPC. The foreseeable convergence demands new abstractions to cope with the increased heterogeneity, so that diverse workload classes can coexist seamlessly on the same infrastructure. In this chapter, we propose a unified storage layer, to enable cloud-native applications to transparently access a wide spectrum of storage solutions, ranging from high-performance filesystems to cloud-based object stores and key–value databases. Allowing software to exploit the best out of the storage services available with no need for manual intervention from users or programmers, simplifies development and execution of workflows, and boosts the overall productivity. This work has been motivated by the requirements of real-world, industry-driven applications running on Kubernetes, the industry standard for cloud infrastructure orchestration.
- Open Access
- 1 January 2016
- Published by LACCEI (Latin American and Caribbean Consortium of Engineering Institutions)
Cryptographic Methods for Deciphering/Identifying Ciphers in MATLAB
- 1 January 2016
- Published by LACCEI (Latin American and Caribbean Consortium of Engineering Institutions)
Open AccessCryptographic Methods for Deciphering/Identifying Ciphers in MATLAB