arXiv subjects

Amin Vahdat

Publications and source records attributed to Amin Vahdat.

17 recordsLinked to original sources

arXiv2026

Quota Marketplace: Dynamic Pricing for Efficient Allocation of ML Training Resources

The escalating demand for Machine Learning (ML) training resources in recent years has resulted in a substantial gap between the high demand and the available supply. Efficient allocation of these scarce and expensive resources is crucial for organizations to maximize their return on investment. Existing resource allocation mechanisms, like Karma [OSDI'23], are designed to guarantee Pareto efficiency and max-min fairness in settings with dynamic (time-varying) user demands, but fail to preserve these key properties in the presence of demands with heterogeneous values. Given the ubiquity and inevitability of heterogeneity in organizational values of different workloads, effective resource allocation policies must accommodate these variations. In this paper, we describe the design, implementation, deployment, and theoretical analysis of Quota Marketplace, a market-based mechanism to efficiently allocate ML training chips (like GPUs), explicitly addressing scenarios with demands of heterogeneous value. We detail the implementation of this mechanism within Google and present metrics that demonstrate its impact. We also discuss many business-critical requirements that the Quota Marketplace handles quite effectively, and document the gains and opportunities it has unlocked. We establish theoretically how this market-based approach achieves the essential properties of Pareto efficiency and max-min fairness by allowing the users to express the value of their workloads and enabling dynamic resource pricing based on supply and demand fluctuations. Ultimately, the market facilitates resource allocation that aligns with organizational priorities.

cs.LG↗

arXiv2025

Measuring the environmental impact of delivering AI at Google Scale

The transformative power of AI is undeniable - but as user adoption accelerates, so does the need to understand and mitigate the environmental impact of AI serving. However, no studies have measured AI serving environmental metrics in a production environment. This paper addresses this gap by proposing and executing a comprehensive methodology for measuring the energy usage, carbon emissions, and water consumption of AI inference workloads in a large-scale, AI production environment. Our approach accounts for the full stack of AI serving infrastructure - including active AI accelerator power, host system energy, idle machine capacity, and data center energy overhead. Through detailed instrumentation of Google's AI infrastructure for serving the Gemini AI assistant, we find the median Gemini Apps text prompt consumes 0.24 Wh of energy - a figure substantially lower than many public estimates. We also show that Google's software efficiency efforts and clean energy procurement have driven a 33x reduction in energy consumption and a 44x reduction in carbon footprint for the median Gemini Apps text prompt over one year. We identify that the median Gemini Apps text prompt uses less energy than watching nine seconds of television (0.24 Wh) and consumes the equivalent of five drops of water (0.26 mL). While these impacts are low compared to other daily activities, reducing the environmental impact of AI serving continues to warrant important attention. Towards this objective, we propose that a comprehensive measurement of AI serving environmental metrics is critical for accurately comparing models, and to properly incentivize efficiency gains across the full AI serving stack.

cs.AI↗

arXiv2025

Accelerating scientific discovery with Co-Scientist

Scientific discovery is driven by scientists generating novel hypotheses for complex problems that undergo rigorous experimental validation. To augment this process, we introduce Co-Scientist, a multi-agent AI system built on Gemini for structured scientific thinking and hypothesis generation. Co-Scientist aims to help scientists discover new original knowledge. Conditioned on their research objectives and prior scientific evidence, it formulates demonstrably novel research hypotheses for experimental verification. The system's design involves agents continuously generating, critiquing and refining hypotheses accelerated by scaling test-time compute. Key contributions include: (1) a multi-agent architecture with an asynchronous task execution framework for flexible compute scaling; (2) a tournament evolution process for self-improving hypotheses generation. Automated evaluations show continued benefits of test-time compute scaling, improving hypothesis quality over time. While general purpose, we focus the validation in three biomedical applications: drug repurposing, novel target discovery, and explaining mechanisms of anti-microbial resistance. Specifically, Co-Scientist helped identify new drug repurposing candidates and synergistic combination therapies for acute myeloid leukemia, which were validated through in vitro experiments. These real-world validations demonstrate the potential of Co-Scientist to accelerate scientific discovery and usher in an era of AI empowered scientists.

cs.AI↗

arXiv2022

Mission Apollo: Landing Optical Circuit Switching at Datacenter Scale

In this paper, we describe Apollo, to the best of our knowledge, the world's first large-scale production deployment of optical circuit switches (OCSes) for datacenter networking. We will first describe the infrastructure challenges and use cases that motivated optical switching inside datacenters. We then delve into the requirements of OCSes for datacenter applications: balancing cost, port count, switching time, and optical performance, which drive design choices and implementation details of our internally developed 3D MEMS-based OCS. To enable the Apollo optical switching layer, we employ circulators to realize bidirectional links through the OCS, effectively doubling the OCS radix. The OCS and circulator design choices were critical for meeting network bandwidth, scale, and cost targets. We review the critical co-design of WDM transceiver technology for these OCS plus circulator-based bidirectional links and their corresponding physical impairments, delivered over four generations/speeds of optical interconnect. Finally, we conclude with thoughts on future directions in hardware development and associated applications.

cs.NI↗

arXiv2021

Gemini: Practical Reconfigurable Datacenter Networks with Topology and Traffic Engineering

To reduce cost, datacenter network operators are exploring blocking network designs. An example of such a design is a "spine-free" form of a Fat-Tree, in which pods directly connect to each other, rather than via spine blocks. To maintain application-perceived performance in the face of dynamic workloads, these new designs must be able to reconfigure routing and the inter-pod topology. Gemini is a system designed to achieve these goals on commodity hardware while reconfiguring the network infrequently, rendering these blocking designs practical enough for deployment in the near future. The key to Gemini is the joint optimization of topology and routing, using as input a robust estimation of future traffic derived from multiple historical traffic matrices. Gemini "hedges" against unpredicted bursts, by spreading these bursts across multiple paths, to minimize packet loss in exchange for a small increase in path lengths. It incorporates a robust decision algorithm to determine when to reconfigure, and whether to use hedging. Data from tens of production fabrics allows us to categorize these as either low-or high-volatility; these categories seem stable. For the former, Gemini finds topologies and routing with near-optimal performance and cost. For the latter, Gemini's use of multi-traffic-matrix optimization and hedging avoids the need for frequent topology reconfiguration, with only marginal increases in path length. As a result, Gemini can support existing workloads on these production fabrics using a spine-free topology that is half the cost of the existing topology on these fabrics.

cs.NI↗

arXiv2018

Eiffel: Efficient and Flexible Software Packet Scheduling

Packet scheduling determines the ordering of packets in a queuing data structure with respect to some ranking function that is mandated by a scheduling policy. It is the core component in many recent innovations to optimize network performance and utilization. Our focus in this paper is on the design and deployment of packet scheduling in software. Software schedulers have several advantages over hardware including shorter development cycle and flexibility in functionality and deployment location. We substantially improve current software packet scheduling performance, while maintaining flexibility, by exploiting underlying features of packet ranking; namely, packet ranks are integers and, at any point in time, fall within a limited range of values. We introduce Eiffel, a novel programmable packet scheduling system. At the core of Eiffel is an integer priority queue based on the Find First Set (FFS) instruction and designed to support a wide range of policies and ranking functions efficiently. As an even more efficient alternative, we also propose a new approximate priority queue that can outperform FFS-based queues for some scenarios. To support flexibility, Eiffel introduces novel programming abstractions to express scheduling policies that cannot be captured by current, state-of-the-art scheduler programming models. We evaluate Eiffel in a variety of settings and in both kernel and userspace deployments. We show that it outperforms state of the art systems by 3-40x in terms of either number of cores utilized for network processing or number of flows given fixed processing capacity.

cs.NI↗

arXiv2015

Quantifying randomness in real networks

Represented as graphs, real networks are intricate combinations of order and disorder. Fixing some of the structural properties of network models to their values observed in real networks, many other properties appear as statistical consequences of these fixed observables, plus randomness in other respects. Here we employ the $dk$-series, a complete set of basic characteristics of the network structure, to study the statistical dependencies between different network properties. We consider six real networks---the Internet, US airport network, human protein interactions, technosocial web of trust, English word network, and an fMRI map of the human brain---and find that many important local and global structural properties of these networks are closely reproduced by $dk$-random graphs whose degree distributions, degree correlations, and clustering are as in the corresponding real network. We discuss important conceptual, methodological, and practical implications of this evaluation of network randomness, and release software to generate $dk$-random graphs.

physics.soc-ph↗

arXiv2015

How small are building blocks of complex networks

Network motifs are small building blocks of complex networks. Statistically significant motifs often perform network-specific functions. However, the precise nature of the connection between motifs and the global structure and function of networks remains elusive. Here we show that the global structure of some real networks is statistically determined by the probability of connections within motifs of size at most 3, once this probability accounts for node degrees. The connectivity profiles of node triples in these networks capture all their local and global properties. This finding impacts methods relying on motif statistical significance, and enriches our understanding of the elementary forces that shape the structure of complex networks.

physics.soc-ph↗

arXiv2014

Programming Protocol-Independent Packet Processors

P4 is a high-level language for programming protocol-independent packet processors. P4 works in conjunction with SDN control protocols like OpenFlow. In its current form, OpenFlow explicitly specifies protocol headers on which it operates. This set has grown from 12 to 41 fields in a few years, increasing the complexity of the specification while still not providing the flexibility to add new headers. In this paper we propose P4 as a strawman proposal for how OpenFlow should evolve in the future. We have three goals: (1) Reconfigurability in the field: Programmers should be able to change the way switches process packets once they are deployed. (2) Protocol independence: Switches should not be tied to any specific network protocols. (3) Target independence: Programmers should be able to describe packet-processing functionality independently of the specifics of the underlying hardware. As an example, we describe how to use P4 to configure a switch to add a new hierarchical label.

cs.NI↗

arXiv2014

Flexible Network Bandwidth and Latency Provisioning in the Datacenter

Predictably sharing the network is critical to achieving high utilization in the datacenter. Past work has focussed on providing bandwidth to endpoints, but often we want to allocate resources among multi-node services. In this paper, we present Parley, which provides service-centric minimum bandwidth guarantees, which can be composed hierarchically. Parley also supports service-centric weighted sharing of bandwidth in excess of these guarantees. Further, we show how to configure these policies so services can get low latencies even at high network load. We evaluate Parley on a multi-tiered oversubscribed network connecting 90 machines, each with a 10Gb/s network interface, and demonstrate that Parley is able to meet its goals.

cs.NI↗

arXiv2010

Hyperbolic Geometry of Complex Networks

We develop a geometric framework to study the structure and function of complex networks. We assume that hyperbolic geometry underlies these networks, and we show that with this assumption, heterogeneous degree distributions and strong clustering in complex networks emerge naturally as simple reflections of the negative curvature and metric property of the underlying hyperbolic geometry. Conversely, we show that if a network has some metric structure, and if the network degree distribution is heterogeneous, then the network has an effective hyperbolic geometry underneath. We then establish a mapping between our geometric framework and statistical mechanics of complex networks. This mapping interprets edges in a network as non-interacting fermions whose energies are hyperbolic distances between nodes, while the auxiliary fields coupled to edges are linear functions of these energies or distances. The geometric network ensemble subsumes the standard configuration model and classical random graphs as two limiting cases with degenerate geometric structures. Finally, we show that targeted transport processes without global topology knowledge, made possible by our geometric framework, are maximally efficient, according to all efficiency measures, in networks with strongest heterogeneity and clustering, and that this efficiency is remarkably robust with respect to even catastrophic disturbances and damages to the network structure.

cond-mat.stat-mech↗

arXiv2010

Greedy Forwarding in Dynamic Scale-Free Networks Embedded in Hyperbolic Metric Spaces

We show that complex (scale-free) network topologies naturally emerge from hyperbolic metric spaces. Hyperbolic geometry facilitates maximally efficient greedy forwarding in these networks. Greedy forwarding is topology-oblivious. Nevertheless, greedy packets find their destinations with 100% probability following almost optimal shortest paths. This remarkable efficiency sustains even in highly dynamic networks. Our findings suggest that forwarding information through complex networks, such as the Internet, is possible without the overhead of existing routing protocols, and may also find practical applications in overlay networks for tasks such as application-level routing, information sharing, and data distribution.

cond-mat.stat-mech↗

arXiv2009

Graph Annotations in Modeling Complex Network Topologies

The coarsest approximation of the structure of a complex network, such as the Internet, is a simple undirected unweighted graph. This approximation, however, loses too much detail. In reality, objects represented by vertices and edges in such a graph possess some non-trivial internal structure that varies across and differentiates among distinct types of links or nodes. In this work, we abstract such additional information as network annotations. We introduce a network topology modeling framework that treats annotations as an extended correlation profile of a network. Assuming we have this profile measured for a given network, we present an algorithm to rescale it in order to construct networks of varying size that still reproduce the original measured annotation profile. Using this methodology, we accurately capture the network properties essential for realistic simulations of network applications and protocols, or any other simulations involving complex network topologies, including modeling and simulation of network evolution. We apply our approach to the Autonomous System (AS) topology of the Internet annotated with business relationships between ASs. This topology captures the large-scale structure of the Internet. In depth understanding of this structure and tools to model it are cornerstones of research on future Internet architectures and designs. We find that our techniques are able to accurately capture the structure of annotation correlations within this topology, thus reproducing a number of its important properties in synthetically-generated random graphs.

cs.NI↗

arXiv2009

Curvature and temperature of complex networks

We show that heterogeneous degree distributions in observed scale-free topologies of complex networks can emerge as a consequence of the exponential expansion of hidden hyperbolic space. Fermi-Dirac statistics provides a physical interpretation of hyperbolic distances as energies of links. The hidden space curvature affects the heterogeneity of the degree distribution, while clustering is a function of temperature. We embed the Internet into the hyperbolic plane, and find a remarkable congruency between the embedding and our hyperbolic model. Besides proving our model realistic, this embedding may be used for routing with only local information, which holds significant promise for improving the performance of Internet routing.

cond-mat.stat-mech↗

arXiv2006

Systematic Topology Analysis and Generation Using Degree Correlations

We present a new, systematic approach for analyzing network topologies. We first introduce the dK-series of probability distributions specifying all degree correlations within d-sized subgraphs of a given graph G. Increasing values of d capture progressively more properties of G at the cost of more complex representation of the probability distribution. Using this series, we can quantitatively measure the distance between two graphs and construct random graphs that accurately reproduce virtually all metrics proposed in the literature. The nature of the dK-series implies that it will also capture any future metrics that may be proposed. Using our approach, we construct graphs for d=0,1,2,3 and demonstrate that these graphs reproduce, with increasing accuracy, important properties of measured and modeled Internet topologies. We find that the d=2 case is sufficient for most practical purposes, while d=3 essentially reconstructs the Internet AS- and router-level topologies exactly. We hope that a systematic method to analyze and synthesize topologies offers a significant improvement to the set of tools available to network topology and protocol researchers.

cs.NI↗

arXiv2005

The Internet AS-Level Topology: Three Data Sources and One Definitive Metric

We calculate an extensive set of characteristics for Internet AS topologies extracted from the three data sources most frequently used by the research community: traceroutes, BGP, and WHOIS. We discover that traceroute and BGP topologies are similar to one another but differ substantially from the WHOIS topology. Among the widely considered metrics, we find that the joint degree distribution appears to fundamentally characterize Internet AS topologies as well as narrowly define values for other important metrics. We discuss the interplay between the specifics of the three data collection mechanisms and the resulting topology views. In particular, we show how the data collection peculiarities explain differences in the resulting joint degree distributions of the respective topologies. Finally, we release to the community the input topology datasets, along with the scripts and output of our calculations. This supplement should enable researchers to validate their models against real data and to make more informed selection of topology data sources for their specific needs.

cs.NI↗

arXiv2005

Lessons from Three Views of the Internet Topology

Network topology plays a vital role in understanding the performance of network applications and protocols. Thus, recently there has been tremendous interest in generating realistic network topologies. Such work must begin with an understanding of existing network topologies, which today typically consists of a relatively small number of data sources. In this paper, we calculate an extensive set of important characteristics of Internet AS-level topologies extracted from the three data sources most frequently used by the research community: traceroutes, BGP, and WHOIS. We find that traceroute and BGP topologies are similar to one another but differ substantially from the WHOIS topology. We discuss the interplay between the properties of the data sources that result from specific data collection mechanisms and the resulting topology views. We find that, among metrics widely considered, the joint degree distribution appears to fundamentally characterize Internet AS-topologies: it narrowly defines values for other important metrics. We also introduce an evaluation criteria for the accuracy of topology generators and verify previous observations that generators solely reproducing degree distributions cannot capture the full spectrum of critical topological characteristics of any of the three topologies. Finally, we release to the community the input topology datasets, along with the scripts and output of our calculations. This supplement should enable researchers to validate their models against real data and to make more informed selection of topology data sources for their specific needs.

cs.NI↗