2026 World Cup: Breaking Records, Spiky Charts and Reliability at Scale
本文为原文前 6,000 字符的节选翻译，完整内容请查看原文。
2026 World Cup: Breaking Records, Spiky Charts and Reliability at Scale
By Syamla Bandla, Vincent Mauge, Michael Stella, Deniz Hastorun, Francois Richard, Javier Frias, Mike McCall Every four years, the World Cup becomes the largest planned stress test of our infrastructure. This year it also set an all-time record! There are a few moments throughout the year, like New Years Eve, that really let us see the collective impact of everything we’ve built — but none are bigger than the quadrennial World Cup tournament. The 2026 World Cup broke records across messaging, video, recommendations, and networking at the same time.
2026年世界杯：打破纪录、激增的图表与大规模可靠性 作者：Syamla Bandla, Vincent Mauge, Michael Stella, Deniz Hastorun, Francois Richard, Javier Frias, Mike McCall。每四年一次的世界杯都会成为我们基础设施最大规模的计划内压力测试。今年，它还创下了历史纪录！一年中只有少数时刻（如新年前夜）能让我们真正看到我们所构建的一切带来的集体影响，但没有哪一个比四年一度的世界杯锦标赛更宏大了。2026年世界杯同时在消息传递、视频、推荐和网络方面打破了纪录。
Some highlights: WhatsApp: More than 30 million messages per second during the July 19 final (Argentina vs. Spain), 1.23x the peak of New Year’s Eve 2026 and an all-time platform record. The previous record had already fallen 12 days earlier, during the Round of 16 match between Argentina and Egypt, at 29 million messages per second. Messenger, Threads, and Instagram Direct: 49 million messages per minute at peak. Instagram: Recommendation system traffic up to 80% higher week-over-week. Final-squad players collectively added +213.6 million activated followers — a 6.6% increase — with combined followings surging from 3.26 billion to 3.47 billion. Live video: Watch time spiked +90% on Instagram and +242% on Facebook, versus the July baseline. Threads: 1.5 billion impressions on tournament-tagged posts, reaching about 15 million people daily and peaking at 25 million on July 6. Facebook: 80 million posts mentioning the World Cup.
部分亮点：WhatsApp：7月19日决赛（阿根廷对阵西班牙）期间，每秒消息量超过3000万条，是2026年新年前夜峰值的1.23倍，创下平台历史纪录。此前的纪录在12天前阿根廷对阵埃及的16强赛中刚刚被打破，当时为每秒2900万条。Messenger、Threads和Instagram Direct：峰值时每分钟消息量达4900万条。Instagram：推荐系统流量周环比增长高达80%。决赛圈球员总共增加了2.136亿活跃粉丝（增长6.6%），总粉丝数从32.6亿激增至34.7亿。直播视频：与7月基准相比，Instagram上的观看时长激增90%，Facebook上激增242%。Threads：带有锦标赛标签的帖子获得了15亿次展示，每天覆盖约1500万人，并在7月6日达到2500万的峰值。Facebook：有8000万篇帖子提到了世界杯。
Traffic Followed the Breaks, Not the Ball Every World Cup match caused surges in traffic. During the final, for example, traffic surged +46% in about four minutes at the end of regular time, and +57% in about 15 minutes at the final whistle. But by lining game events up against global egress traffic we discovered a striking pattern – surges in sharing were primarily happening during breaks (hydration breaks, halftime, extra time, the final whistle, the awards ceremony). The end-of-regulation whistle produced the single biggest peak of the tournament. It wasn’t the gameplay that moved our charts. It was the conversations after it. The world came to Meta to watch the game and to share what it made them feel. Every capacity decision below follows from that one pattern.
流量跟随比赛间歇，而非比赛进程。每场世界杯比赛都会导致流量激增。例如，在决赛中，常规时间结束时流量在约四分钟内激增了46%，终场哨响时在约15分钟内激增了57%。但通过将比赛事件与全球出口流量进行对比，我们发现了一个显著的模式——分享激增主要发生在比赛间歇（补水时间、中场休息、加时赛、终场哨响、颁奖典礼）。常规时间结束的哨声产生了锦标赛中最大的单次峰值。推动我们图表变化的不是比赛过程，而是赛后的讨论。全世界来到Meta观看比赛并分享他们的感受。以下每一项容量决策都遵循这一模式。
Layered Systems, No Manual Intervention Those spikes were absorbed by systems built for exactly this. Taiji balanced load across datacenters, Katran and Proxygen handled L4/L7 traffic distribution, our global load balancing system routed users to the best-positioned edge, and AutoScaling provisioned capacity on demand. Region-Specific Capacity Months before kickoff, Meta’s edge teams analyzed network capacity and worked closely with ISP partners so that upgrades landed where they would most improve quality of experience for end users. That targeted pre-build is now permanent capacity that will keep serving demand long after the tournament.
分层系统，无需人工干预。这些峰值被专门为此构建的系统所吸收。Taiji在数据中心之间平衡负载，Katran和Proxygen处理L4/L7流量分配，我们的全球负载均衡系统将用户引导至位置最佳的边缘节点，而自动扩容（AutoScaling）则按需配置容量。特定区域容量：在开赛前几个月，Meta的边缘团队分析了网络容量并与ISP合作伙伴密切合作，确保升级部署在最能改善终端用户体验的地方。这种有针对性的预先构建现已成为永久容量，将在锦标赛结束后长期满足需求。
Instagram: Preparing for 39 Consecutive Days of Surges For Instagram, the tournament meant 39 straight days of traffic spikes as billions of fans checked scores, posted Stories, and flooded Feed during breaks. Machine learning inference, Feed serving, Stories, Explore, and data infrastructure all had to absorb it without degrading the experience. During the final, halftime Instagram Feed traffic rose 76% above the previous day in the minute the match ended, and stayed about 21% above for the following two hours. Our approach rested on four things, deployed incrementally as the tournament moved from group stage to final. Capacity buffers: Our products went into the tournament with sufficient capacity, which meant real buffers were available to allocate. We pre-provisioned significant additional capacity across all serving tiers so that spikes never had to wait on autoscaling to catch up. These dedicated buffers were our single most effective lever of the tournament. Country-level graceful degradation: For high-risk matches, we shed non-critical background load in the top-traffic nations (Brazil, Argentina, India, Indonesia, Spain) to protect core functionality. Cascade prevention: So spikes at the serving layer couldn’t propagate into downstream数据 systems. Early warning signals These alerted us before we approached capacity thresholds, rather than after. All of this, including the phased deployment timeline and the degradation activation criteria, is now a reusable runbook for future large-scale events like the 2028 Olympics.
Instagram：为连续39天的流量激增做准备。对于Instagram来说，锦标赛意味着连续39天的流量高峰，数十亿球迷在比赛间歇查看比分、发布快拍（Stories）并涌入信息流（Feed）。机器学习推理、信息流服务、快拍、探索（Explore）以及数据基础设施都必须在不降低体验的情况下吸收这些流量。在决赛中，比赛结束后的那一分钟，Instagram信息流流量比前一天上升了76%，并在接下来的两个小时内保持在约21%的增幅。我们的方法基于四点，随着锦标赛从小组赛进入决赛阶段逐步部署。容量缓冲：我们的产品在进入锦标赛时拥有充足的容量，这意味着有实际的缓冲可供分配。我们在所有服务层预先配置了大量额外容量，使峰值流量无需等待自动扩容。这些专用缓冲是我们锦标赛中最有效的手段。国家级优雅降级：对于高风险比赛，我们在流量最高的国家（巴西、阿根廷、印度、印度尼西亚、西班牙）削减非关键的后台负载，以保护核心功能。级联预防：防止服务层的峰值传播到下游数据系统。早期预警信号：这些信号在我们接近容量阈值之前就向我们发出警报，而不是事后。所有这些，包括分阶段部署时间表和降级激活标准，现在都成为了未来大型活动（如2028年奥运会）的可复用操作手册。
WhatsApp: 8.4 Billion Messages in Five Minutes Goals and the final whistle drove WhatsApp spikes lasting four to five minutes at a time. In its single highest-traffic five-minute window, WhatsApp delivered nearly 8.4 billion messages without interruption. That came from months of preparation. More than 100 internal services were load-tested and resized ahead of the tournament, then retested, revalidated, and resized as the bracket narrowed and demand climbed. Two things we’d normally do, we didn’t. We didn’t shift load between regions. For past high-traffic events, our default has been to spread demand across multiple regions, letting capacity in quieter parts of the world absorb the spike. That works when peaks land hours apart — a global New Year’s Eve, for example. This event was different. Every region peaked in the same instant, so there was no idle capacity to borrow. Instead of deriving regional demand from a global multiplier, we forecast each region’s peak on its own terms. We didn’t freeze. The other standard playbook move for a major event is to pause continuous deployment and hold off on riskier operations like disaster recovery exercises. This time we skipped the freeze entirely, on the strength of years of reliability enhancement. Teams shipped on their normal cadence throughout the entire event window, including during the peak moment itself. We Stay Ready Year-Round Enabling connection in moments like this doesn’t just happen. Powering every message sent, every follower, and every Ree
WhatsApp：五分钟内发送84亿条消息。进球和终场哨声引发了WhatsApp持续四到五分钟的流量激增。在流量最高的五分钟窗口内，WhatsApp不间断地传送了近84亿条消息。这源于数月的准备工作。锦标赛前，超过100项内部服务进行了负载测试和调整，随后随着赛程推进和需求攀升，又进行了重新测试、验证和调整。我们没有做两件通常会做的事。我们没有在区域间转移负载。对于过去的高流量事件，我们的默认做法是将需求分散到多个区域，让世界上较安静地区的容量来吸收峰值。这在峰值相隔数小时出现时（例如全球新年前夜）是有效的。但这次活动不同。每个区域在同一瞬间达到峰值，因此没有闲置容量可以借用。我们没有通过全球乘数来推导区域需求，而是根据每个区域自身的情况预测其峰值。我们没有冻结部署。大型活动的另一个标准操作是暂停持续部署，并推迟灾难恢复演习等高风险操作。这一次，凭借多年来可靠性增强的实力，我们完全跳过了冻结。在整个活动期间，包括在峰值时刻，团队都保持正常的发布节奏。我们全年保持就绪。在这样的时刻实现连接并非偶然。为每一条发送的消息、每一个关注者以及每一个Ree提供支持。