arXiv subjects

Konstantina Mellou

Publications and source records attributed to Konstantina Mellou.

8 recordsLinked to original sources

arXiv2026

Online Rack Placement in Large-Scale Data Centers: Online Sampling Optimization and Deployment

This paper optimizes the configuration of large-scale data centers toward cost-effective, reliable and sustainable cloud supply chains. The problem involves placing incoming racks of servers within a data center to maximize demand coverage given space, power and cooling restrictions. We formulate an online integer optimization model to support rack placement decisions. We propose a tractable online sampling optimization (OSO) approach to multi-stage stochastic optimization, which approximates unknown parameters with a sample path and re-optimizes decisions dynamically. We prove that OSO achieves a strong competitive ratio in canonical online resource allocation problems and sublinear regret in the online batched bin packing problem. Theoretical and computational results show it can outperform mean-based certainty-equivalent resolving heuristics. Our algorithm has been packaged into a software solution deployed across Microsoft's data centers, contributing an interactive decision-making process at the human-machine interface. Using deployment data, econometric tests suggest that adoption of the solution has a negative and statistically significant impact on power stranding, estimated at 1-3 percentage point. At the scale of cloud computing, these improvements in data center performance result in significant cost savings and environmental benefits.

math.OC↗

arXiv2026

Online Scheduling for LLM Inference with KV Cache Constraints

Large Language Model (LLM) inference, where a trained model generates text one word at a time in response to user prompts, is a computationally intensive process requiring efficient scheduling to optimize latency and resource utilization. A key challenge in LLM inference is the management of the Key-Value (KV) cache, which reduces redundant computations but introduces memory constraints. In this work, we model LLM inference with KV cache constraints theoretically and propose a novel batching and scheduling algorithm that minimizes inference latency while effectively managing the KV cache's memory. More specifically, we make the following contributions. First, to evaluate the performance of online algorithms for scheduling in LLM inference, we introduce a hindsight optimal benchmark, formulated as an integer program that computes the minimum total inference latency under full future information. Second, we prove that no deterministic online algorithm can achieve a constant competitive ratio when the arrival process is arbitrary. Third, motivated by the computational intractability of solving the integer program at scale, we propose a polynomial-time online scheduling algorithm and show that under certain conditions it can achieve a constant competitive ratio. We also demonstrate our algorithm's strong empirical performance by comparing it to the hindsight optimal in a synthetic dataset. Finally, we conduct empirical evaluations on a real-world public LLM inference dataset, simulating the Llama2-70B model on A100 GPUs, and show that our algorithm significantly outperforms the benchmark algorithms. Overall, our results offer a path toward more sustainable and cost-effective LLM deployment.

cs.LG↗

arXiv2026

OptiMind: Teaching LLMs to Think Like Optimization Experts

Mathematical programming -- the task of expressing operations and decision-making problems in precise mathematical language -- is fundamental across domains, yet remains a skill-intensive process requiring operations research expertise. Recent advances in large language models for complex reasoning have spurred interest in automating this task, translating natural language into executable optimization models. Current approaches, however, achieve limited accuracy, hindered by scarce and noisy training data without leveraging domain knowledge. In this work, we systematically integrate optimization expertise to improve formulation accuracy for mixed-integer linear programming, a key family of mathematical programs. Our OptiMind framework leverages semi-automated, class-based error analysis to guide both training and inference, explicitly preventing common mistakes within each optimization class. Our resulting fine-tuned LLM significantly improves formulation accuracy by 20.7% across multiple optimization benchmarks, with consistent gains under test-time scaling methods such as self-consistency and multi-turn feedback, enabling further progress toward robust LLM-assisted optimization formulation.

cs.LG↗

arXiv2025

Large Language Models for Supply Chain Decisions

Supply Chain Management requires addressing a variety of complex decision-making challenges, from sourcing strategies to planning and execution. Over the last few decades, advances in computation and information technologies have enabled the transition from manual, intuition and experience-based decision-making, into more automated and data-driven decisions using a variety of tools that apply optimization techniques. These techniques use mathematical methods to improve decision-making. Unfortunately, business planners and executives still need to spend considerable time and effort to (i) understand and explain the recommendations coming out of these technologies; (ii) analyze various scenarios and answer what-if questions; and (iii) update the mathematical models used in these tools to reflect current business environments. Addressing these challenges requires involving data science teams and/or the technology providers to explain results or make the necessary changes in the technology and hence significantly slows down decision making. Motivated by the recent advances in Large Language Models (LLMs), we report how this disruptive technology can democratize supply chain technology - namely, facilitate the understanding of tools' outcomes, as well as the interaction with supply chain tools without human-in-the-loop. Specifically, we report how we apply LLMs to address the three challenges described above, thus substantially reducing the time to decision from days and weeks to minutes and hours as well as dramatically increasing planners' and executives' productivity and impact.

cs.AI↗

arXiv2024

The Power of Migrations in Dynamic Bin Packing

In the Dynamic Bin Packing problem, $n$ items arrive and depart the system in an online manner, and the goal is to maintain a good packing throughout. We consider the objective of minimizing the total active time, i.e., the sum of the number of open bins over all times. An important tool for maintaining an efficient packing in many applications is the use of migrations; e.g., transferring computing jobs across different machines. However, there are large gaps in our understanding of the approximability of dynamic bin packing with migrations. Prior work has covered the power of no migrations and $> n$ migrations, but we ask the question: What is the power of limited ($\leq n$) migrations? Our first result is a dichotomy between no migrations and linear migrations: Using a sublinear number of migrations is asymptotically equivalent to doing zero migrations, where the competitive ratio grows with $μ$, the ratio of the largest to smallest item duration. On the other hand, we prove that for every $α\in (0,1]$, there is an algorithm that does $\approx αn$ migrations and achieves competitive ratio $\approx 1/α$ (in particular, independent of $μ$); we also show that this tradeoff is essentially best possible. This fills in the gap between zero migrations and $> n$ migrations in Dynamic Bin Packing. Finally, in light of the above impossibility results, we introduce a new model that more directly captures the impact of migrations. Instead of limiting the number of migrations, each migration adds a delay of $C$ time units to the item's duration; this commonly appears in settings where a blackout or set-up time is required before the item can restart its execution in the new bin. In this new model, we prove a $O(\min (\sqrt{C}, μ))$-approximation, and an almost matching lower bound.

cs.DS↗

arXiv2023

Large Language Models for Supply Chain Optimization

Supply chain operations traditionally involve a variety of complex decision making problems. Over the last few decades, supply chains greatly benefited from advances in computation, which allowed the transition from manual processing to automation and cost-effective optimization. Nonetheless, business operators still need to spend substantial efforts in explaining and interpreting the optimization outcomes to stakeholders. Motivated by the recent advances in Large Language Models (LLMs), we study how this disruptive technology can help bridge the gap between supply chain automation and human comprehension and trust thereof. We design OptiGuide -- a framework that accepts as input queries in plain text, and outputs insights about the underlying optimization outcomes. Our framework does not forgo the state-of-the-art combinatorial optimization technology, but rather leverages it to quantitatively answer what-if scenarios (e.g., how would the cost change if we used supplier B instead of supplier A for a given demand?). Importantly, our design does not require sending proprietary data over to LLMs, which can be a privacy concern in some circumstances. We demonstrate the effectiveness of our framework on a real server placement scenario within Microsoft's cloud supply chain. Along the way, we develop a general evaluation benchmark, which can be used to evaluate the accuracy of the LLM output in other scenarios.

cs.AI↗

arXiv2022

Online Demand Scheduling with Failovers

Motivated by cloud computing applications, we study the problem of how to optimally deploy new hardware subject to both power and robustness constraints. To model the situation observed in large-scale data centers, we introduce the Online Demand Scheduling with Failover problem. There are $m$ identical devices with capacity constraints. Demands come one-by-one and, to be robust against a device failure, need to be assigned to a pair of devices. When a device fails (in a failover scenario), each demand assigned to it is rerouted to its paired device (which may now run at increased capacity). The goal is to assign demands to the devices to maximize the total utilization subject to both the normal capacity constraints as well as these novel failover constraints. These latter constraints introduce new decision tradeoffs not present in classic assignment problems such as the Multiple Knapsack problem and AdWords. In the worst-case model, we design a deterministic $\approx \frac{1}{2}$-competitive algorithm, and show this is essentially tight. To circumvent this constant-factor loss, which in the context of big cloud providers represents substantial capital losses, we consider the stochastic arrival model, where all demands come i.i.d. from an unknown distribution. In this model we design an algorithm that achieves a sub-linear additive regret (i.e. as OPT or $m$ increases, the multiplicative competitive ratio goes to $1$). This requires a combination of different techniques, including a configuration LP with a non-trivial post-processing step and an online monotone matching procedure introduced by Rhee and Talagrand.

cs.DS↗

arXiv2020

Online Virtual Machine Allocation with Predictions

The cloud computing industry has grown rapidly over the last decade, and with this growth there is a significant increase in demand for compute resources. Demand is manifested in the form of Virtual Machine (VM) requests, which need to be assigned to physical machines in a way that minimizes resource fragmentation and efficiently utilizes the available machines. This problem can be modeled as a dynamic version of the bin packing problem with the objective of minimizing the total usage time of the bins (physical machines). Earlier works on dynamic bin packing assumed that no knowledge is available to the scheduler and later works studied models in which lifetime/duration of each "item" (VM in our context) is available to the scheduler. This extra information was shown to improve exponentially the achievable competitive ratio. Motivated by advances in Machine Learning that provide good estimates of workload characteristics, this paper studies the effect of having extra information regarding future (total) demand. In the cloud context, since demand is an aggregate over many VM requests, it can be predicted with high accuracy (e.g., using historical data). We show that the competitive factor can be dramatically improved by using this additional information; in some cases, we achieve constant competitiveness, or even a competitive factor that approaches 1. Along the way, we design new offline algorithms with improved approximation ratios for the dynamic bin-packing problem.

cs.DS↗