arXiv subjects

Junaid Ahmed Khan

Publications and source records attributed to Junaid Ahmed Khan.

6 recordsLinked to original sources

arXiv2026

Symbolic Separation: Grounding Deep Agents in Knowledge Graphs for Trustworthy Operational Data Analytics

Generative AI promises natural language access to the massive numerical telemetry of data centers and Industry 4.0 installations, yet text-to-query and tool-using agents stay unreliable: even frontier models answer little more than half of real-world database questions, and far fewer of the multi-step, operational ones, because the LLM must compose how heterogeneous sources relate and hallucinates the relations, not just the fields. We propose symbolic separation: a deep agent reasons freely but may act on data only through an ontology-constrained Virtual Knowledge Graph with deterministic pre-execution validation. Unlike a tool API's interface contract, this domain-semantic contract turns a complex question into one validated graph traversal instead of LLM-inferred joins. Instantiated as the Neurosymbolic Deep Analyst and evaluated on 49.9 TB of superconputer telemetry against a rigid workflow and a non-symbolic ablation, it raises end-to-end task success from 43% to 86%, prevents silent data-integrity errors that no syntactic check catches, and cuts token cost by 2.4x, letting a smaller on-premise model outperform a larger one.

cs.AI↗

arXiv2025

From Data Center IoT Telemetry to Data Analytics Chatbots -- Virtual Knowledge Graph is All You Need

Industry 5.0 demands IoT systems that support seamless human-machine collaboration, yet current IoT data analysis requires deep domain, deployment, and query expertise. We show that combining Large Language Models (LLMs) with Knowledge Graphs (KGs) enables natural language access to heterogeneous IoT data. Focusing on data center IoT telemetry, we introduce a rule-based Virtual Knowledge Graph (VKG) construction process and an on-premise LLM inference service to create an end-to-end Data Analytics (DA) chatbot. Our system dynamically generates VKGs per query and translates user input into SPARQL, achieving 92.5% accuracy (vs. 25% for LLM-to-NoSQL) while reducing latency by 85% (20.36s to 3.03s) and keeping VKG sizes under 179 MiB. This work demonstrates that VKG-powered LLM interfaces deliver accurate, low-latency, and relationship-aware access to large-scale telemetry, bridging the gap between users and complex IoT systems in Industry 5.0.

cs.DC↗

arXiv2025

A Unified Ontology for Scalable Knowledge Graph-Driven Operational Data Analytics in High-Performance Computing Systems

Modern high-performance computing (HPC) systems generate massive volumes of heterogeneous telemetry data from millions of sensors monitoring compute, memory, power, cooling, and storage subsystems. As HPC infrastructures scale to support increasingly complex workloads-including generative AI-the need for efficient, reliable, and interoperable telemetry analysis becomes critical. Operational Data Analytics (ODA) has emerged to address these demands; however, the reliance on schema-less storage solutions limits data accessibility and semantic integration. Ontologies and knowledge graphs (KG) provide an effective way to enable efficient and expressive data querying by capturing domain semantics, but they face challenges such as significant storage overhead and the limited applicability of existing ontologies, which are often tailored to specific HPC systems only. In this paper, we present the first unified ontology for ODA in HPC systems, designed to enable semantic interoperability across heterogeneous data centers. Our ontology models telemetry data from the two largest publicly available ODA datasets-M100 (Cineca, Italy) and F-DATA (Fugaku, Japan)-within a single data model. The ontology is validated through 36 competency questions reflecting real-world stakeholder requirements, and we introduce modeling optimizations that reduce knowledge graph (KG) storage overhead by up to 38.84% compared to a previous approach, with an additional 26.82% reduction depending on the desired deployment configuration. This work paves the way for scalable ODA KGs and supports not only analysis within individual systems, but also cross-system analysis across heterogeneous HPC systems.

cs.DC↗

arXiv2019

Reversing The Meaning of Node Connectivity for Content Placement in Networks of Caches

It is a widely accepted heuristic in content caching to place the most popular content at the nodes that are the best connected. The other common heuristic is somewhat contradictory, as it places the most popular content at the edge, at the caching nodes nearest the users. We contend that neither policy is best suited for caching content in a network and propose a simple alternative that places the most popular content at the least connected node. Namely, we populate content first at the nodes that have the lowest graph centrality over the network topology. Here, we provide an analytical study of this policy over some simple topologies that are tractable, namely regular grids and trees. Our mathematical results demonstrate that placing popular content at the least connected nodes outperforms the aforementioned alternatives in typical conditions.

cs.NI↗

arXiv2017

Offloading Content with Self-organizing Mobile Fogs

Mobile users in an urban environment access content on the internet from different locations. It is challenging for the current service providers to cope with the increasing content demand from a large number of collocated mobile users. In-network caching to offload content at nodes closer to users alleviate the issue, though efficient cache management is required to find out who should cache what, when and where in an urban environment, given nodes limited computing, communication and caching resources. To address this, we first define a novel relation between content popularity and availability in the network and investigate a node's eligibility to cache content based on its urban reachability. We then allow nodes to self-organize into mobile fogs to increase the distributed cache and maximize content availability in a cost-effective manner. However, to cater rational nodes, we propose a coalition game for the nodes to offer a maximum "virtual cache" assuming a monetary reward is paid to them by the service/content provider. Nodes are allowed to merge into different spatio-temporal coalitions in order to increase the distributed cache size at the network edge. Results obtained through simulations using realistic urban mobility trace validate the performance of our caching system showing a ratio of 60-85% of cache hits compared to the 30-40% obtained by the existing schemes and 10% in case of no coalition.

cs.NI↗

arXiv2017

A Content-based Centrality Metric for Collaborative Caching in Information-Centric Fogs

Information-Centric Fog Computing enables a multitude of nodes near the end-users to provide storage, communication, and computing, rather than in the cloud. In a fog network, nodes connect with each other directly to get content locally whenever possible. As the topology of the network directly influences the nodes' connectivity, there has been some work to compute the graph centrality of each node within that network topology. The centrality is then used to distinguish nodes in the fog network, or to prioritize some nodes over others to participate in the caching fog. We argue that, for an Information-Centric Fog Computing approach, graph centrality is not an appropriate metric. Indeed, a node with low connectivity that caches a lot of content may provide a very valuable role in the network. To capture this, we introduce acontent-based centrality (CBC) metric which takes into account how well a node is connected to the content the network is delivering, rather than to the other nodes in the network. To illustrate the validity of considering content-based centrality, we use this new metric for a collaborative caching algorithm. We compare the performance of the proposed collaborative caching with typical centrality based, non-centrality based, and non-collaborative caching mechanisms. Our simulation implements CBC on three instances of large scale realistic network topology comprising 2,896 nodes with three content replication levels. Results shows that CBC outperforms benchmark caching schemes and yields a roughly 3x improvement for the average cache hit rate.

cs.NI↗