T2TRG Chong, Song Internet-Draft KAIST Intended status: Standards Track Jang, Hyeonjoon Expires: December 15, 2021 KAIST October 2020 Low End-to-End Latency Content Caching in Wireless Network Clouds draft-chong-t2trg-llwnc-00 Abstract In this document, we consider the content caching design without requiring historical content access information or content popularity profiles in a hierarchical cellular network architecture. Our design aims to dynamically select caching locations for different contents where caching locations can be content servers, cloud units (CUs), and base stations (BSs). Our design objective is to support as high content request rates as possible while maintaining the finite service time. Status of this Memo This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at http://datatracker.ietf.org/drafts/current/. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." This Internet-Draft will expire on December 15, 2021. Chong & Jang Expires December 15, 2021 [Page 1] Internet-Draft Wireless Cloud Content Caching October 2020 Copyright Notice Copyright (c) 2020 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . .. . . . . . 2 2. Main Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.1. System Model . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.2. Hybrid Content Caching Design . . . . . . . . . . . .. . . . . 4 3. IANA Considerations . . . . . . .. . . . . . . . . . . . . . . 5 4. Security Considerations . . . . . . . . . . . . . . . . . . . 5 5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5.1. Normative References . . . . . . . . . . . . . . . . . . . . . 5 5.2. Informative References . . . . . .. . . . . . . . . . . . . . 5 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . . 6 1. Introduction With the rapidly increasing mobile video traffic, both backhaul and fronthaul networks connecting the internet with the mobile core and edge networks such as base stations (BSs) (see Fig. 1) become more and more congested. Since popular contents are more frequently requested by end users, we can reduce the end-to-end latency of video content services and backhaul/fronthaul traffic loads by placing popular contents at locations closer to the end users. Content popularity profile, which captures its average access frequency from end users, can be spatio-temporally varying in wireless networks due to user mobility or social interaction between mobile users. Moreover, the massive deployment of small cells such as femto-cells in the cellular networks increases the spatial granularity which renders the real-time estimation of the content popularity in each BS more challenging. Meanwhile, edge-centric technologies (e.g., edge computing/ caching and distributed Self-Organizing Network (SON)) and cloud-centric technologies (e.g., cloud computing/caching and Cloud Radio Access Network (C-RAN)) have been devised to support the latency-critical applications and to address the huge amount of workloads in the cloud servers, respectively. Recently, hybrid network architecture and operations which adaptively exploit the edge-centric and cloud-centric natures of the wireless network environments, has been proposed. Chong & Jang Expires December 15, 2021 [Page 2] Internet-Draft Wireless Cloud Content Caching October 2020 2. Main Idea In this document we consider the caching design in a general wireless network architecture where each group of a small number of BSs is connected to a cloud unit(CU) where individual BSs and the CU have content caching repositories (see Fig. 1). The caching design must adaptively determine the content caching locations at BSs and CUs depending on the network dynamics to support as high average content request rates as possible with finite content service time. To address this design while not requiring information on historical content access or content popularity profile, we employ the Lyapunov optimization method [a] for which the short-term max-weight problem derived from the Lyapunov drift must be optimized in the slot-by-slot basis. Because the max-weight problem is NP-hard and difficult to tackle [b] due to the coupling between CU and BS caching decisions, we propose an approximation algorithm which achieves the constant approximation ratio to the optimal weight by exploiting the submodularity of the slot-by-slot objective function and the structure of hierarchical content caching networks. +------------------+ +-------+ +--+ End-to-End +------+ | Original Content |---------| Cloud |--------|BS| Path(case1) | End | | Servers | Backhaul| Unit | Frount +--+ <==========>| | +------------------+ +-------+ haul | users| | <=================================================>+------+ Backhaul | End-to-End Path(case3) | +-------+ Frouthaul +--+ +------+ | Cloud |----------- |BS| ------------| End | | | +--+ | users| | Unit |<===========================>+------+ +-------+ End-to-End Path(case2) Figure 1: Network Architecture 2.1 System Model Fig. 1 illustrates the hierarchical network architecture considered in the dynamic caching design. We consider a video content set where the file size of different contents is assumed to be the same and equal to s (in bits). Each video content file is assumed to be split into multiple chunks, and we assume that a file can be recovered at the destination even if not all of the chunks for the file are successfully delivered [c]. We assume there are E cloud units (CUs) and N base stations (BSs) in the system. All contents are saved in their own original content servers distributed throughout the internet. We consider a time-slotted system with equal-size time slots of duration Δt and the time slots are indexed as t = 0, 1, ...2. Moreover, caching control decisions are made in the slot-by slot basis. In each time slot t, the amount of data from content requested by end users associated with BSs and CUs is independent and identically distributed over time slots. Chong & Jang Expires December 15, 2021 [Page 3] Internet-Draft Wireless Cloud Content Caching October 2020 When content f is requested from an end user, the requested content is transmitted from its closest caching location among the associated BS, CU and original content server which has the corresponding content placed by a deployed caching strategy. Hereafter, we call this closest network node as a source node of content f. Hence, the end-to-end path of content f could vary as the choice of the source node changes. (Figure 1.) To capture the dynamics of content requests, services, and the corresponding content service time, we introduce virtual queues for each BS where the virtual queue backlog for content f, i.e., Q_f(t), evolves over time as follows: Q_f(t+1) = [Q_f(t) - r_f(t) + A_f(t)]^+ , for every CU and associated BS, content(f), where [x]^+ = max(x,0). In the above, we define A_f(t) and r_f(t) as the amount of data from content f requested by end users associated with BSs and the amount of served data of content f at BSs, respectively. The amount of served data r_f(t) from each virtual queue during time slot t depends on the caching decision, i.e., to cache content f at CU or BS. The average virtual queue backlog of content f indirectly captures the average end-to-end latency of content f. 2.2. Hybrid Content Caching Design The hybrid content caching algorithm should (i) achieve low average end-to-end latency by stabilizing virtual queues if the content request rate vector is within the capacity region and (ii) support as high average content request rates as possible. For any content request rate vector inside the capacity region,all virtual queues must be stable, i.e., supremum of the following value t-1 (1/r)* Σ {expectation of sum of Q_f(r) w.r.t every f, BS and CU} r=0 should be bounded as t goes to infinity. To develop such a caching algorithm, we could employ the Lyapunov optimization method [a]. Toward this end, we define Lyapunov function and Lyapunov drift function w.r.t. Q_f(t) and derive an upper bound of the Lyapunov drift function using the queueing dynamics of the virtual queues. Then the content caching algorithm can be developed by minimizing the upper bound of the Lyapunov drift function in each time slot. Chong & Jang Expires December 15, 2021 [Page 4] Internet-Draft Wireless Cloud Content Caching October 2020 3. IANA Considerations There are no IANA considerations related to this document. 4. Security Considerations There are no security considerations related to this document. 5. References 5.1. Normative References [a] M. Neely, “Stochastic network optimization with application to communication and queueing systems,” Synthesis Lectures on Communication Networks, pp. 1–211, 2010. [b] R. Hemmecke, M. Koppe, J. Lee, and R. Weismantel, 50 Years of Integer Programming 1958-2008. Springer, 2009. [c] F. Pantisano, M. Bennis, W. Saad, and M. Debbah, “Match to Cache :Joint user association and backhaul allocation in cache-aware small cell networks,” in Proc. of IEEE ICC, London, UK, Jun. 2015, pp. 3082–3087. 5.2. Informative References Acknowledgements This work was supported by Institute for Information & communications Technology Promotion(IITP) grant funded by the Korea government(MSIT) (No.2015-0-00557, Resilient/Fault-Tolerant Autonomic Networking Based on Physicality, Relationship and Service Semantic of IoT Devices) Chong & Jang Expires December 15, 2021 [Page 5] Internet-Draft Wireless Cloud Content Caching October 2020 Authors' Addresses Song Chong The Graduate School of Artificial Intelligence, Korea Advanced Institute of Science and Technology(KAIST) Daejeon, South Korea Phone: +82 (0)42 350 3473 Email: songchong@kaist.edu Hyeonjoon Jang Electrical Engineering Department, Korea Advanced Institute of Science and Technology(KAIST) Daejeon, South Korea Phone: +82 (0)42 350 5473 Email: thefelix@kaist.ac.kr Sewoong Lee Electrical Engineering Department, Korea Advanced Institute of Science and Technology(KAIST) Daejeon, South Korea Phone: +82 (0)42 350 5473 Email: dltpdnd21@kaist.ac.kr Chong & Jang Expires December 15, 2021 [Page 6]