The Demand for Online at All Costs

Photo by Leon Seibert on Unsplash

Italian version available here

News from the past months: "Libero and Alice's emails are not working. They have not been working for days and will resume as soon as possible." Everyone is outraged - I understand - but in fact, seeing the level of advertising, spam, and the like, I would say it was clear that the two services were no longer being managed "impeccably".

News appeared in the same period, Microsoft had problems, so Outlook, Teams, etc. had serious disservices for hours.

Even Facebook, Whatsapp, and Instagram had serious problems in the past. And these are all services that lose millions of euros for each minute of downtime.

Yet today, in the interconnected world, a disruption in connectivity or service is seen as a tragedy.

I've been managing servers for so long that a child born when I put my first services online could have graduated and had children of their own. My first server was archived by the Wayback Machine in 2002, but it's at least a year older. Some colleagues I work with daily were born after my first server.

Yet today, like never before, a stop of a service (even outside of business hours, even if planned, even if the service is not needed at that moment) has become unacceptable.

Every morning a system administrator wakes up and knows that they have to race against time to patch, check, and reboot services before a vulnerability can hit their system. Every morning a user wakes up and will have to use the service just in the few seconds in which the system administrator is rebooting it. Even if the system administrator had warned that they would reboot the service at that time, causing a "few minutes" of interruption. Even if the system administrator got up at 4 in the morning to do it. Even if the user, at 4 in the morning on Sunday, usually sleeps.

The Network is made of interconnected components and only parts of it are controllable. It can happen that something jumps: a service provider, a backbone, an external dns. We have to learn to accept that something, sometimes, may not be fully efficient.

When you have a physical store, there can be a power outage, water, gas, a flood, road work... why can't what's online be idle for a minute every 6 months, if announced? Or an hour every year, if there is an unforeseen problem?

The promises of the "cloud" have led everyone to believe that 24/7 always and anyway exists. But no, it actually doesn't exist and the more complex the infrastructure, the more parts can break. And the promises of "always online" are often drowned in terms and conditions with very limited liability in case of breach.

Better five minutes offline today than an attack tomorrow, with the relative risk of leakage of personal data.

Written and posted by: Stefano Marinelli