arXiv subjects

Ferdi Kossmann

Publications and source records attributed to Ferdi Kossmann.

5 recordsLinked to original sources

arXiv2025

ECCO: Leveraging Cross-Camera Correlations for Efficient Live Video Continuous Learning

Recent advances in video analytics address real-time data drift by continuously retraining specialized, lightweight DNN models for individual cameras. However, the current practice of retraining a separate model for each camera suffers from high compute and communication costs, making it unscalable. We present ECCO, a new video analytics framework designed for resource-efficient continuous learning. The key insight is that the data drift, which necessitates model retraining, often shows temporal and spatial correlations across nearby cameras. By identifying cameras that experience similar drift and retraining a shared model for them, ECCO can substantially reduce the associated compute and communication costs. Specifically, ECCO introduces: (i) a lightweight grouping algorithm that dynamically forms and updates camera groups; (ii) a GPU allocator that dynamically assigns GPU resources across different groups to improve retraining accuracy and ensure fairness; and (iii) a transmission controller at each camera that configures frame sampling and coordinates bandwidth sharing with other cameras based on its assigned GPU resources. We conducted extensive evaluations on three distinctive datasets for two vision tasks. Compared to leading baselines, ECCO improves retraining accuracy by 6.7%-18.1% using the same compute and communication resources, or supports 3.3 times more concurrent cameras at the same accuracy.

cs.DC↗

arXiv2025

KramaBench: A Benchmark for AI Systems on Data-to-Insight Pipelines over Data Lakes

Discovering insights from a real-world data lake potentially containing unclean, semi-structured, and unstructured data requires a variety of data processing tasks, ranging from extraction and cleaning to integration, analysis, and modeling. This process often also demands domain knowledge and project-specific insight. While AI models have shown remarkable results in reasoning and code generation, their abilities to design and execute complex pipelines that solve these data-lake-to-insight challenges remain unclear. We introduce KramaBench which consists of 104 manually curated and solved challenges spanning 1700 files, 24 data sources, and 6 domains. KramaBench focuses on testing the end-to-end capabilities of AI systems to solve challenges which require automated orchestration of different data tasks. KramaBench also features a comprehensive evaluation framework assessing the pipeline design and individual data task implementation abilities of AI systems. We evaluate 8 LLMs using our single-agent reference framework DS-Guru, alongside both open- and closed-source single- and multi-agent systems, and find that while current agentic systems may handle isolated data-science tasks and generate plausible draft pipelines, they struggle with producing working end-to-end pipelines. On KramaBench, the best system reaches only 55% end-to-end accuracy in the full data-lake setting. Even with perfect retrieval, the accuracy tops out at 62%. Leading LLMs can identify up to 42% of important data tasks but can only fully implement 20% of individual data tasks. Our code, reference framework, and data are available at https://github.com/mitdbg/KramaBench.

cs.DB↗

arXiv2024

Is the GPU Half-Empty or Half-Full? Practical Scheduling Techniques for LLMs

Serving systems for Large Language Models (LLMs) improve throughput by processing several requests concurrently. However, multiplexing hardware resources between concurrent requests involves non-trivial scheduling decisions. Practical serving systems typically implement these decisions at two levels: First, a load balancer routes requests to different servers which each hold a replica of the LLM. Then, on each server, an engine-level scheduler decides when to run a request, or when to queue or preempt it. Improved scheduling policies may benefit a wide range of LLM deployments and can often be implemented as "drop-in replacements" to a system's current policy. In this work, we survey scheduling techniques from the literature and from practical serving systems. We find that schedulers from the literature often achieve good performance but introduce significant complexity. In contrast, schedulers in practical deployments often leave easy performance gains on the table but are easy to implement, deploy and configure. This finding motivates us to introduce two new scheduling techniques, which are both easy to implement, and outperform current techniques on production workload traces.

cs.LG↗

arXiv2024

Blueprinting the Cloud: Unifying and Automatically Optimizing Cloud Data Infrastructures with BRAD -- Extended Version

Modern organizations manage their data with a wide variety of specialized cloud database engines (e.g., Aurora, BigQuery, etc.). However, designing and managing such infrastructures is hard. Developers must consider many possible designs with non-obvious performance consequences; moreover, current software abstractions tightly couple applications to specific systems (e.g., with engine-specific clients), making it difficult to change after initial deployment. A better solution would virtualize cloud data management, allowing developers to declaratively specify their workload requirements and rely on automated solutions to design and manage the physical realization. In this paper, we present a technique called blueprint planning that achieves this vision. The key idea is to project data infrastructure design decisions into a unified design space (blueprints). We then systematically search over candidate blueprints using cost-based optimization, leveraging learned models to predict the utility of a blueprint on the workload. We use this technique to build BRAD, the first cloud data virtualization system. BRAD users issue queries to a single SQL interface that can be backed by multiple cloud database services. BRAD automatically selects the most suitable engine for each query, provisions and manages resources to minimize costs, and evolves the infrastructure to adapt to workload shifts. Our evaluation shows that BRAD meet user-defined performance targets and improve cost-savings by 1.6-13x compared to serverless auto-scaling or HTAP systems.

cs.DB↗

arXiv2024

CascadeServe: Unlocking Model Cascades for Inference Serving

Machine learning (ML) models are increasingly deployed to production, calling for efficient inference serving systems. Efficient inference serving is complicated by two challenges: (i) ML models incur high computational costs, and (ii) the request arrival rates of practical applications have frequent, high, and sudden variations which make it hard to correctly provision hardware. Model cascades are positioned to tackle both of these challenges, as they (i) save work while maintaining accuracy, and (ii) expose a high-resolution trade-off between work and accuracy, allowing for fine-grained adjustments to request arrival rates. Despite their potential, model cascades haven't been used inside an online serving system. This comes with its own set of challenges, including workload adaption, model replication onto hardware, inference scheduling, request batching, and more. In this work, we propose CascadeServe, which automates and optimizes end-to-end inference serving with cascades. CascadeServe operates in an offline and online phase. In the offline phase, the system pre-computes a gear plan that specifies how to serve inferences online. In the online phase, the gear plan allows the system to serve inferences while making near-optimal adaptations to the query load at negligible decision overheads. We find that CascadeServe saves 2-3x in cost across a wide spectrum of the latency-accuracy space when compared to state-of-the-art baselines on different workloads.

cs.DC↗