VNF-Cache: An In-Network Key-Value Store Cache Based on Network Function Virtualization
Bruno E. Farias, José Flauzino, Elias P. Duarte Jr · 2025
Read VNF-Cache: An In-Network Key-Value Store Cache Based on Network Function Virtualization by Bruno E. Farias, José Flauzino, Elias P. Duarte Jr (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2512-19964.txt.
Introduction
Bruno Farias, José Flauzino, and Elias Duarte put the cache inside the network itself. Key-value databases serving distant clients suffer latency on every lookup, and classic caching near the server does little for geography. VNF-Cache is a caching service built on network function virtualization with computing in the network, an IETF-standardizing technology that runs arbitrary services on the data path.
The service intercepts packets in flight, processes and stores values, and answers clients directly whenever the answer is already held. A proof-of-concept deployment spanned geographically dispersed servers in Brazil, the United States, and Japan. Experiments there showed significant cuts in response time and clear gains in requests processed per second, evidence for in-network caching as a practical tool.
Catalog record
| CREATOR | Bruno E. Farias, José Flauzino, Elias P. Duarte Jr |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2512-19964
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
VNF-Cache: An In-Network Key-Value Store Cache Based on Network Function Virtualization Bruno E. Farias, José Flauzino, Elias P. Duarte Jr Abstract With the exponential growth of the amount of data available on the Internet, optimizing the response time and resource usage for data access becomes essential. Caches are an effective solution that brings data closer to clients, eliminating repetitive requests to servers. This paper presents VNF-Cache, a caching service for geographically remote key-value databases. VNF-Cache is an NFV-COIN (Network Function Virtualization-Computing In The Network) service, a technology undergoing standardization by the IETF that enables the implementation of arbitrary services directly in the network. VNF-Cache intercepts network packets, processes, stores, and sends values directly to clients when possible. Through a proof-of-concept implementation and experiments conducted with geographically dispersed servers in Brazil, the United States, and Japan, significant reductions in response time and increases in the number of requests processed per second were observed. 1 Introduction Caches are a fundamental mechanism for improving the efficiency of data access over the Internet. This paper introduces VNF-Cache, a caching service for key-value databases built with Network Functions Virtualization (NFV) technology. NFV enables the implementation of network functions as software based on virtualization technologies [ 3 ] . Given the wide variety of network functions that can be deployed using NFV, the NFV-MANO (NFV Management and Orchestration) reference architecture was established to standardize their management and ensure interoperability across NFV ecosystems [ 4 ] . This architecture defines the lifecycle management of Virtualized Network Functions (VNFs) and the computational resources they require [ 23 ] . Recently, the NFV-COIN (NFV - COmputing In the Network) architecture has been proposed to support the deployment of arbitrary and innovative services directly within the network, through the COIN (COmputing In-the-Network) paradigm [ 35 ] . NFV-COIN, is currently undergoing standardization within the IETF, which has already published two drafts defining two foundation components: one outlining the problem statement [ 21 ] and another specifying the architectural framework [ 20 ] . The VNF-Cache proposed in this work is designed as an NFV-COIN service. Its purpose is to optimize client access to geographically distant key-value databases by placing data closer to end users. By reducing the distance that requests and responses must traverse, VNF-Cache lowers access latency and increases the throughput of processed requests. Additionally, by avoiding unnecessary end-to-end communication with the backend server, the service contributes to a more efficient utilization of network resources. To provide key-value caching capabilities, VNF-Cache must be located at an intermediate point in the network between clients and the corresponding database servers. Its operation begins by identifying and filtering relevant traffic, after which it processes and stores the values associated with keys requested by clients. When cached data is available, VNF-Cache responds directly to the client; otherwise, the request is forwarded to the server, and the corresponding response is intercepted and stored, when applicable, to serve future requests. The VNF-Cache architecture is described in detail, along with a proof-of-concept prototype. An empirical evaluation was conducted across three experimental scenarios: (A) in the first scenario a client, VNF-Cache, and key-value database server are all located in close proximity; (B) in the second scenario, the client and VNF-Cache are close to each other, while the server is geographically distant; and finally (C) in the third scenario the client, VNF-Cache, and server all located distant to each other. The remote instances of VNF-Cache and the database servers were deployed on Amazon Elastic Compute Cloud (Amazon EC2), a cloud computing service from Amazon Web Services (AWS) 1 1 1 https://aws.amazon.com/pt/ . Virtual machines were instantiated in Curitiba and São Paulo (Brazil), Ohio (USA), and Tokyo (Japan). The results show that, in scenarios where clients and servers are geographically distant, introducing VNF-Cache can substantially reduce request response times and increase the throughput of processed requests. The remainder of this work is organized as follows. Section 2 presents an overview of NFV and NFV-COIN technologies, as well as related work. Section 3 describes the VNF-Cache, detailing its architecture and the prototype that was implemented. Section 4 reports the experimental results. Finally, Section 5 provides concluding remarks and outlines directions for future work. 2 NFV, NFV-COIN & Related Work Network Functions Virtualization, or NFV, is a technology for implementing network services using virtualization, such as Virtual Machines (VMs) and containers, which can run on off-the-shelf hardware. A wide variety of network services can be implemented with NFV, such as routers, Virtual Private Networks (VPNs), traffic analyzers, firewalls, Content Delivery Networks (CDNs), among others [ 3 ] . Before NFV technology became available, the single alternative was to have those services implemented in specialized hardware, which is a substantially less flexible approach. NFV has changed the global communication networks scenario, by broadening the scope of vendors of network funtions and services, which are made available through global marketplaces [ 1 ] . In order to establish standards and develop an architecture for managing the life cycle of virtualized network functions and services is the MANagement and Orchestration (MANO) model [ 4 ] . MANO emerged as an architecture to enable interoperability between NFV systems. To this end, the architecture encompasses not only the lifecycle management of VNFs and but also of the computational resources they use. A very large number of NFV platforms based on MANO are currently available [ 25 , 5 , 6 ] . Note that a key component of an NFV platform is the VNF execution platform [ 14 , 15 ] , which is responsible for the actual execution of the network functions [ 18 , 26 ] . The NFV-MANO architecture is basically composed of three fundamental and interdependent blocks: the Virtualized Infrastructure Manager (VIM), the VNF Manager (VNFM), and the NFV Orchestrator (NFVO), described next. The VIMs are responsible for controlling and managing the physical infrastructure and resources available for virtualization. A single VIM can be responsible for managing only one type of computing resource at a time (such as processing, memory, or network), or several types of resources simultaneously. It must be able to efficiently distribute the resources made available to an NFVI (or NFV Infrastructure) in order to meet the needs of each VNF without impacting the execution of the others. The NFVI is the set of physical hardware together with the virtualization layer that runs the VNFs. The VNF Manager (VNFM) is responsible for managing the lifecycle of VNF instances; that is, this module is responsible for instantiating and removing instances, performing configurations and updates, as well as monitoring and scaling functions [ 31 ] . A single VNFM is responsible for managing several virtualized network functions of different types, the EM (Element Management) instrumentalizes the VNF for the purpose of management and interactions with the VNFM [ 9 ] . The NFV Orchestrator (NFVO) has two main functions: managing the NFVI across multiple existing VIMs and coordinating the lifecycle of network services. The NFVO is the component that enables the composition of VNFs that for a Service Function Chain (SFC) to create complex services [ 16 ] . It is possible to combine multiple different VNFs, each solving some specific problem, in different topologies [ 8 ] . Besides SFC