Zoom: Architected for Reliability
Introduction
Zoom delivers an AI-first Unified Communications as a Service (UCaaS) and Contact Center as a Service (CCaaS) through a single, integrated platform. The suite includes Zoom Meetings; Zoom Webinars and Zoom Events; Zoom Team Chat; Zoom Rooms and Workspace Reservation; Zoom Phone; Zoom Contact Center; Zoom Docs; Zoom Mail and Calendar Service; Zoom Whiteboard; Zoom Virtual Agent; Zoom Revenue Accelerator; and more—built to work together across desktop, mobile, browser, and conference rooms, including standards-based SIP/H.323 systems via the Conference Room Connector.
This document details the innovations and architecture behind Zoom’s best-in-class services and products, explaining how we deliver a consistent, secure, and manageable experience for end users and IT across UCaaS and CCaaS.
Built from Experience
Zoom’s cloud-first vision helps organizations move beyond the cost and complexity of legacy on-premises systems. Rather than bolting features onto aging stacks, we’ve invested in full-stack engineering across clients, media services, and room systems to optimize end-to-end quality, resiliency, and usability as one cohesive platform—anchored by a single-client strategy with a common codebase across platforms that takes advantage of each OS’s native UX.
That cloud-first approach shows up in the Zoom Workspace app: a single, modern application that gives users centralized access to Meetings, Team Chat, Phone, Contact Center, Events, Webinars, Rooms, Whiteboard, Docs, and more. Access is governed by licensing, so IT can turn products and features on or off without deploying a patchwork of separate installers or managing conflicting client versions.
For IT, the model reduces overhead—fewer servers to buy, rack, and patch; fewer bespoke integrations to maintain; and no need for entire teams dedicated to keeping on-prem infrastructure running. Capacity, updates, and security improvements are delivered from the cloud, so admins can focus on policy, adoption, and outcomes rather than maintenance.
Innovation Makes the Difference
Zoom’s unique approach to the way we’ve architected our service means that Zoom has features that other solutions just can’t provide, at a scale, and with quality that is unmatched.
AI Companion: Built into Meetings, Team Chat, Phone, Contact Center, Docs, and more to summarize, draft, and surface next steps right where work happens—no extra apps or context-switching.
Smart Gallery & Intelligent Director: Multi-stream, multi-camera Rooms that give everyone in the room a proper “seat” on screen.
Continuous Meeting Chat: Meeting chat rolls into a Team Chat channel so links and files don’t vanish when the meeting ends, supporting continuous, asynchronous collaboration.
Proximity Share & Echo Prevention: Walk into a room, share wirelessly in a click; nearby laptops auto-mute to prevent feedback.
Media Resiliency: Conversations stay intelligible under rough networks via adaptive bitrate, codec layers, and fast transport fallback.
Zoom Mesh: A native eCDN that redistributes the same meeting or webinar media inside your network using peer-to-peer technology to reduce bandwidth consumption during large sessions.
Zoom Node: Run selected workloads on-premises with cloud control and lifecycle management, including local survivability for Meetings and Phone.
Post-Quantum E2EE: End-to-end encryption designed to resist future decryption attacks for meetings and 1:1 phone calls.
Data-Routing Controls: Choose which regions you approve for media traversal.
Quality and Reliability
The following sections discuss Zoom’s design approach to enable a best-in-class real-time media experience.
Zoom Meetings, Webinars, and Events
The foundation of our architecture is an intelligent transport layer that selects the best path based on real network conditions and policies. UDP is used when available, with seamless fallback to TCP/TLS (including HTTPS/443) in more restrictive environments. Screen sharing uses Reliable UDP for smooth motion whenever possible, with automatic fallback if needed. When connecting to a meeting, webinar, or event, the Zoom Workplace app is steered by geolocation to the nearest available resources to minimize latency. Where organizational policies or performance require, traffic can traverse Zoom’s global backbone via dedicated links to different data centers.
The next layer is a reactive quality-of-service layer that adapts to real-time network and device conditions. It monitors bandwidth, packet loss, latency, and jitter, and also collects local CPU usage, memory, and network I/O. These signals inform the upper layers to take the appropriate adaptive actions to maintain quality and reliability.
Zoom’s custom, adaptive codec at the session layer is tuned for real-time performance. Surrounding algorithms continuously optimize frame rate and resolution to match device and network conditions. To provide the best possible experience as conditions allow, Zoom uses multiple simultaneous streams, and the Zoom Workplace app dynamically selects the most appropriate layer. Thanks to efficient compression, sessions remain usable with packet loss up to ~45%; in those cases, audio is prioritized over video to keep conversations clear. The multi-stream approach also adjusts bandwidth to each participant’s ability to send and receive audio and video.
Zoom’s distributed conferencing layer uses subscription-based switching without server-side transcoding or mixing. Traditional services often transcode and mix streams, adding CPU and memory overhead; our switching approach is designed to reduce resource use and scale efficiently. Participants are routed by geolocation to the nearest data center and assigned to the least-loaded server; when attendees are co-located, they can be grouped on the same server for efficiency. The architecture supports flexible on-premises and hybrid deployments and provides cascaded traffic paths for large enterprises.
The Meeting Server is our MMR (multi-media router), and MMRs are grouped in a “Meeting Zone.” Zone Controllers manage all of the MMRs and report their status to the Global Cloud Controller for each Meeting Zone. The Meeting Zones are duplicated for each data center with the exact same architecture and we can easily add more zones on-the-fly for added capacity in each region. The three layers (the MMR, Zoom Controller, and Global Cloud Controller) are used to balance resources in different locations. If just two participants are in a meeting, Zoom will utilize peer-to-peer connections for excellent speed and reliability. All of this allows Zoom to maintain meeting services availability of 99.9% uptime and provide a dependable video conferencing experience.
Zoom Phone
Zoom Phone was built in the cloud and for the cloud, using the proven audio quality adaptations already available in our Zoom Meetings and Webinars products. Zoom’s architecture has redundancy and resiliency built in, resulting in a highly-available solution that can scale to meet even the largest organization’s needs.
Zoom Phone utilizes intelligent transport mechanisms for reliable connectivity across diverse network environments. For media, it uses SRTP over UDP when available, with seamless fallback to TCP/TLS, along with adaptive bitrate and packet-loss mitigation to preserve audio quality under poor conditions. For connectivity, calls route to the nearest available SIP Zone to minimize latency, with automatic failover across data centers when needed.
Zoom has redundant Session Border Controllers (SBCs) in each of its data centers that secure client and carrier communications. These carrier-grade SBCs facilitate easy access for a broad range of organizations, from the smallest of customers to global enterprises. Load Balancers redirect SIP-based communications to Zoom’s Call Switches to evenly distribute the call volume. This distribution enables a smooth experience for users, even during peak registration and busy call hours.
Call Switches are the core call control of Zoom Phone. These scalable components not only support base PBX functionality, but also facilitate telemetry data to the Zoom Phone Dashboard and enable features like elevating phone calls to Zoom Meetings.
The Zoom Workplace app leverages proprietary logic to monitor the app’s bandwidth, packet loss, latency, and jitter, while also collecting the app’s CPU usage, memory, and network I/O. This technology actively monitors calls and makes real-time adjustments to overcome poor network conditions to provide superior call quality and reliability for various network environments and different devices.
The Zoom Phone Dashboard captures real-time and historical quality of service data, as well as usage and adoption metrics, call logs, and metrics related to nomadic emergency services. Zoom automatically rates each call with a Mean Opinion Score (MOS), allowing IT administrators to track the performance of all calls traversing the network and isolate potential network-related issues.
Zoom Contact Center
Similar to Zoom Phone, Zoom Contact Center’s active-active architecture helps provide resiliency and redundancy across its omni-channel environment. Building on proven architectures from Zoom Meetings and Phone services, Contact Center seamlessly integrates real-time communication capabilities with web functionality, creating an intuitive interface that benefits both your team of agents and your customers.
Each data center features two identical, interconnected Session Initiation Protocol (SIP) zones. SIP zones allow voice calls to be made and received over the internet, and each zone is equipped with dedicated hardware and services for independent sustainability. Within a data center, a load balancer evenly distributes calls between both SIP zones. Calls are distributed among a cluster of call switches, which are responsible for various functions such as call routing, setup, and teardown.
From the call switches, calls connect to an SBC within each zone, which either connects to Zoom’s underlying network of providers or a customer-provided carrier for PSTN routing until the call reaches its final destination. SBCs, load balancers, and call switches are supplemented with redundant hardware on standby for resiliency.
Voice in Zoom Contact Center benefits from the same client adaptation and transport behavior as Zoom Phone—using UDP when available with fallback to TCP/TLS—so audio stays clear even on constrained networks. For video and digital channels, the service leverages Zoom’s real-time media backbone and cloud resiliency for consistent quality and failover.
Zoom Node for Hybrid & On-Premises Deployments
Zoom Node is a cloud-managed, hybrid platform that links your data center resources with Zoom’s cloud, enabling you to deploy Zoom Service Modules (workloads) on-premises and manage them from the Zoom web portal. Two modules stand out for resiliency: Zoom Meetings Hybrid and Zoom Phone Local Survivability (ZPLS).
Zoom Meetings Hybrid routes real-time media for internal participants through your local Node while signaling, admin, and policy remain in the Zoom cloud, and it includes a local survivability option if Zoom data centers become unreachable. Internal users get low-latency paths and reduced internet egress, while external participants can join over the cloud as usual. Accounts can also add Recording Connector for local capture when required, with automatic cloud failover if local capacity isn’t available.
Zoom Phone Local Survivability (ZPLS) keeps core calling available at a site if the WAN or provider path goes down. Phones register locally, internal extension calling continues, and, when configured, critical outbound calling can be maintained through an on-premises PSTN path. When connectivity returns, service automatically resumes normal cloud operation.
Global Distributed Data Centers and Redundancy
Zoom has brokers and communications servers distributed among multiple interconnected data centers across the globe. We are constantly evaluating our data centers and Internet service providers (ISPs) to optimize performance for our customers in regards to bandwidth, latency, and disaster recovery isolation. Our data centers are situated in secure co-location facilities that are ISP carrier neutral and provide physical security, redundant power, and simultaneous access to top-tier ISPs and peering partners. We also have certain deployments on public cloud providers where regional or capacity requirements warrant. Our co-location facilities are built with fault-tolerant architecture with full redundancy and rapid failover capability. Zoom dynamically load balances the communications servers to automatically move new sessions to the data center that has the best response time.
Capacity
Zoom maintains 50% excess capacity in all aspects of our infrastructure to accommodate our growing business and to meet peak usage requirements. We are confident in our ability to provide service and scale based on our current and future customer needs.
Disaster Recovery
Zoom regularly validates its disaster recovery capabilities through no less than annual testing conducted across its suite of products. Because Zoom uses active-active architecture with redundancy in each scope or layer, there is minimal risk of interrupted service. All data centers are also fully redundant with power, cooling, and network carriers. Multiple power feeds, fiber links, backup generators and battery systems provide reliability.
Conclusion
Zoom’s innovative approach in delivery and features provides a consistently superior video conferencing experience. End users appreciate the innovative features, ease of use, reliability and incredible video and audio clarity. And IT managers are assured that the solution is globally available and designed to scale with security and dependability.
Last updated
Was this helpful?