Multicloud

Multicloud means a single company relying on several public cloud providers at once. On June 8, 2021, a configuration change made by a Fastly customer triggered a latent software bug, and about 85% of the network returned errors for close to an hour. Sites hosted on AWS (Amazon Web Services), Google Cloud or Azure went down together, because they all ran through that same content delivery network. That is the first problem with multicloud: diversity is measured at the hosting layer, and dependency sits higher up. DNS (Domain Name System), the CDN (content delivery network), the identity provider, the security agent pushed to every machine all cut across providers, and July 19, 2024 reminded the whole market of it. The second problem is more mundane. At many companies that call themselves multicloud, each application lives with a single provider: payroll with one, data analytics with the other. Nothing fails over if one goes down; the loss is just split. An active-active architecture across two clouds does exist, at some banks and some online retailers, but it is expensive and rare. For the underwriter, the "multicloud" box on the questionnaire is worth little without two answers: which critical applications can run elsewhere, and how fast; which upstream services are shared. Those answers feed accumulation modeling, which works provider by provider and where a CDN common to a whole portfolio can weigh as much as a cloud region. They also help read dependent business interruption cover, often sublimited depending on whether the provider is named or not.