Third-party services (Keynote, Gomez; cheaper PingDom, SiteUptime, Alertra) ping your pages from worldwide machines and build availability/latency dashboards. As "objective" outsiders, their numbers can underwrite SLAs. But before trusting them, understand what they actually measure: - Are they simulating real human users? - Do they cache objects like a normal browser — why or why not? - Can you separate network-transfer time from server time, in aggregate and per object? - Can you tell a real failure from a geographic-network or measurement artifact? Only if the monitor requests pages the way your users do are the numbers trustworthy. A second caveat: these metrics often land on an executive dashboard. Transparency helps justify spend, but a non-technical, obsessive audience will fixate on anomalies — be ready to explain what each blip means. --- *Source: [[The Art of Capacity Planning]] (John Allspaw, O'Reilly 2008) — Ch 2 — Setting Goals for Capacity*