- Internet disruptions around the world increased 63% in March compared to January, according to ThousandEyes' Internet Performance Report based on the company's platform and publicly disclosed outages. Disruptions continued into June, increasing 45% compared to January.
- Cloud providers had approximately 400 outages between January and July, compared to about 4,500 outages in internet service provider (ISP) networks, including telecom, transit and broadband providers. With outages increasing 150%, cloud providers experienced an "unusual rise" in North America this year. ThousandEyes defines outages as "terminal events with 100% packet loss."
- Content delivery network (CDN) and domain name system providers had four outages in North America but held onto "good availability," according to the report. "A small number of outages could conceivably have more of an impact on network availability than a large number of outages if they were severe enough."
The internet was not built for resilience. It's an interconnected network built on the border gateway protocol (BGP), with autonomous systems leading to it. BGP was originally made to uphold trust between ISPs and entities, without doubting the integrity of flowing information.
Now bad actors are capitalizing on the BGP's ability to share routing. If outages result from massive upticks in traffic, such as remote work and stay-at-home orders, they can ripple through businesses, no matter their connection to the internet or an ISP.
While ISPs had more frequent outages, the overall business impact is relative, depending on the type of business and its region. ISP outages in North America typically occur outside of business hours, and weekend disruptions are less likely than during the weekdays. The report suggests the discrepancies in when outages occurred are because there are "fewer staff resources available to make network changes and/or lower overall traffic levels."
When "network duress" is underway, companies will likely experience, in order, the following:
- Elevated latency and jitters
- Variability in interpacket gaps, leading to drops in traffic
- Packet retransmissions, leading to increased delays and congestion
For North America, packet losses peaked in May, reaching 5%, compared to 1% in EMEA. However, the packet loss, even at its spikes, isn't strong enough to user experience disruption.
Companies with chaos engineers self-inflict outages. Companies that experienced their own outages, unrelated to internet health, were already lagging in reliability. Reliability correlates directly to scalability and uptime.
Cloud computing will always be a company's best bet for scaling, and its newer technology frees it from "the technical debt of longstanding operators," compared to ISPs, according to the report.
ISPs are more available than the cloud for some, but the cloud showed less degradation in the first half of 2020. Internet performance and reliability will impact what features SaaS companies can enable.