**Parent Topic**: [[Software/README]] ## Even the Best Vendors Fail "Just as the most reliable datacenters go down from time to time, you can expect even the best third-party vendors to have issues now and then." A single vendor is a single point of failure outside your walls. ## Collateral Damage from Shared Vendors Christian's cautionary tale: in 2006, anti-spam startup Blue Security's escalating war with spammers ended when they launched "a massive DDoS attack against Blue's DNS provider, UltraDNS." Because UltraDNS hosted many high-profile customers, "the battle resulted in significant collateral damage" — innocent bystanders like Tucows and Six Apart "were taken offline through no fault of their own." Similar 2009 attacks caused secondary outages at Amazon and Salesforce.com. ## The Rule "You can't put all your faith in a single vendor, especially if the vendor is hosting multiple companies. Someone else might get attacked, and through no fault of your own, your site goes down." The solution: "always maintain multi-vendor solutions, particularly when they rely on shared resources. Go in with the assumption that at some point, one of the vendors will suffer an outage, and you can just switch to the other." The external-dependency form of [[Monitor What You Don't Run]] — your reliability is bounded by your vendors', so don't depend on just one. ## Related - [[Single Point of Failure (SPOF)]] — the general concept this rule instantiates; that note's mitigation table lists "Single DNS provider → Multiple DNS providers," the same shared-vendor risk as the UltraDNS collateral-damage case here. - [[Freelancing Risk Reframe — Distributed vs Concentrated Risk]] — same "don't concentrate dependency in one entity" logic applied to career income instead of infrastructure vendors. - [[Model-Harness Decoupling as Vendor Risk Mitigation]] — "date the models, marry the harness": the AI-architecture version of not trusting a single provider. - [[Trade Concentration Vulnerability]] — the geopolitical/maritime-trade version of the same fragility pattern: throughput concentrated through one chokepoint fails catastrophically when that point is disrupted. --- *Source: [[Web Operations]] (Allspaw & Robbins, O'Reilly 2010) — Ch 17 — Things That Go Bump in the Night*