|6 min read

The World Went Remote Overnight

COVID-19 just forced the largest remote work experiment in human history, and cloud infrastructure is the only reason it is working at all

Five days ago, the WHO declared COVID-19 a pandemic. Three days ago, the President declared a national emergency. Two days ago, my entire organization, thousands of employees at a major entertainment company, went remote. Not in a planned, phased transition over six months with change management workshops and ergonomic assessments. Overnight. Everyone got sent home.

This is happening everywhere. Offices are empty. Schools are closed. The entire white-collar workforce of the developed world is attempting to do their jobs from kitchen tables, spare bedrooms, and couches. And somehow, mostly, it is working.

The Infrastructure That Made This Possible

If this pandemic had hit in 2010, remote work at this scale would have been impossible. The infrastructure simply did not exist. VPN concentrators would have collapsed under the load. On-premises Exchange servers would have buckled. Video conferencing at consumer scale was unreliable at best.

In 2020, we have a different technology stack. And it is being stress-tested in real time.

Cloud computing is the foundation. Our workloads run on AWS. Our CI/CD pipelines, our container orchestration, our data platforms, our monitoring systems: all of it runs in the cloud. When everyone went home, the infrastructure did not move. The applications did not change. The developers just opened their laptops in different locations and kept working.

This is the promise of cloud computing made tangible. Location independence. The compute, storage, and networking layer does not care where the human is sitting. The API endpoints are the same. The deployment pipelines are the same. The Git repositories are the same.

SaaS collaboration tools are carrying the human communication layer. Slack, Zoom, Google Workspace, Confluence, Jira: the entire productivity stack is browser-based and cloud-hosted. There is no server in a closet somewhere that needs to be online for people to communicate. The capacity scales with demand, at least in theory.

In practice, Zoom is struggling. The sudden surge in video conferencing demand is exposing capacity limits. Calls are dropping. Audio quality is degrading. Zoom's engineering team is probably not sleeping much right now. But the fundamental architecture, cloud-based, horizontally scalable, globally distributed, is correct. The question is whether they provisioned enough headroom for a demand spike that nobody modeled.

VPN is the Bottleneck

The one piece of infrastructure that was not ready is VPN. Most enterprises, including ours, sized their VPN infrastructure for a fraction of the workforce connecting remotely at any given time. The assumption was that maybe 20-30% of employees would need VPN access simultaneously, primarily for accessing internal tools and databases.

When 100% of the workforce went remote in 48 hours, the VPN concentrators hit their license limits and throughput ceilings. Connections were being throttled or dropped. The experience was degraded enough that people were actively looking for workarounds.

This exposed a deeper architectural question: why do we need VPN at all? For applications that live in the cloud, VPN is routing traffic through the corporate network for no functional reason other than to satisfy a security model designed for a perimeter-based world. The zero-trust networking model, where identity and device posture determine access rather than network location, has been discussed in security circles for years. BeyondCorp at Google proved it works at scale. But most enterprises, including ours, have been slow to adopt it.

The pandemic is going to accelerate zero-trust adoption faster than any vendor pitch deck ever could. When VPN is the thing preventing your workforce from being productive, the business case for alternatives becomes very clear very quickly.

What Changed in Our Engineering Workflow

For our platform engineering team specifically, the transition has been smoother than I expected. Our workflow was already heavily asynchronous:

  • Code reviews happen in pull requests with written comments. No in-person whiteboarding required.
  • Deployments are automated through CI/CD pipelines. Nobody needs to be in a specific location to ship code.
  • Incident response was already running through PagerDuty and Slack. War rooms were already virtual.
  • Architecture discussions are the one area where remote is noticeably worse. Complex system design benefits from a whiteboard and real-time back-and-forth in ways that Zoom does not fully replicate.

The teams that are struggling most are the ones whose workflows depended on physical proximity: hardware labs, production studios, operations centers with specialized equipment. Knowledge work is portable. Physical work is not.

Scaling Observations

Some numbers from the first week:

Our AWS spend has not changed significantly. The workload profile is roughly the same; the requests are just coming from different IP ranges. This validates the architectural decision to run stateless services behind load balancers with auto-scaling groups. The system does not care about the source of the request.

Slack message volume is up roughly 40%. People are compensating for the loss of hallway conversations with more Slack messages. This is predictable but has implications for notification fatigue and the difficulty of context-switching in a high-interrupt environment.

CI/CD pipeline throughput has actually increased slightly. My hypothesis is that developers, freed from meetings and commute time, are shipping more code. Whether this is sustainable or a temporary burst remains to be seen.

The Uncomfortable Questions

This experiment is going to force some uncomfortable conversations when it is over.

If the entire company can function remotely, what is the office for? The answer is probably nuanced: collaboration, culture, mentoring, the kinds of unstructured interaction that produce serendipitous ideas. But the default assumption that knowledge workers must be physically present to be productive has been shattered.

If cloud infrastructure can support the entire workforce operating remotely, why are we still running anything on-premises? Every on-prem system that required VPN access to function has been a pain point this week. Every cloud-native system has been fine. The migration calculus just shifted.

If VPN is a bottleneck and zero-trust networking works at Google's scale, why are we not investing in it? The security team has been talking about zero-trust for two years. The budget conversation just got a lot easier.

The Human Side

The technology is working. The humans are struggling. Working from home with children who are also home because schools are closed, with a global pandemic generating constant anxiety, with no clear timeline for when normalcy returns: this is not the remote work utopia that digital nomad blogs describe.

I have been checking in with my team daily, not to track productivity, but to make sure people are okay. Some are thriving. Some are overwhelmed. The extroverts are climbing the walls. The introverts are quietly relieved. Everyone is uncertain.

What I keep telling my team is this: we will figure out the work. The pipelines will run. The deployments will ship. The systems will stay up. Focus on taking care of yourselves and your families. The infrastructure can handle the load. We need to make sure the people can too.

The world went remote overnight. The technology held. Now we need to make sure the humans do.

Share: