OneNym.idDiscuss deployment
English

THE BOARDROOM FIREWALL / Chapter 9

Living in the Cloud

Moving to the cloud is like renting a floor in a well-run tower instead of owning the building. This chapter explains where the landlord's job ends and yours begins, and which decisions about data location, outages and unapproved tools belong to leadership.

This page summarises the chapter. The full text is in the book.

Renting rather than owning

The cloud is other companies' computers in other companies' data centres, rented rather than owned. It comes in degrees. Renting raw computing and storage (infrastructure as a service) is like taking an empty floor that your own team fits out. Renting a platform gives developers a ready environment. Renting finished software (software as a service, or SaaS), such as email or accounting tools, means almost everything is provided. Most companies, including the book's fictional Meridian Foods, use all three.

The more finished the service, the more the provider is responsible for and the less your team can see or change. That shifts what your security depends on: with raw capacity, mostly your own team's skill; with finished software, mostly the provider's skill and how carefully your people use the settings.

Shared responsibility: landlord and tenant

Every major provider publishes a version of the shared-responsibility model. The provider secures the cloud itself: data centres, hardware, core network and the software that runs the service. The customer secures what happens in it: who can access accounts, how data is configured and shared, which security settings are switched on and how staff use the service. The line moves with the type of service, but some share always stays with the customer.

Industry reports find that many serious cloud incidents trace back to the customer's side: storage left public by mistake, accounts without multi-factor authentication, excessive permissions and unchanged default settings. The most dangerous assumption is that using a big provider means security has been outsourced. When Meridian's chief executive asked for a one-page split of responsibilities for its five most important cloud services, the exercise revealed two tasks nobody owned: reviewing administrator access to the finance system and checking that the customer app was actually being backed up.

Where your data lives

Data residency is where data is physically stored; data sovereignty is the wider idea that data is subject to the laws of that country, and sometimes of the provider's home country. Leaders should care for three reasons: data-protection laws such as Europe's GDPR and Japan's APPI place conditions on cross-border transfers, large customers increasingly ask where their data will be held, and data in a jurisdiction can be subject to legal requests from its authorities. Major providers usually let you choose a storage region, but backups, support access and newer features may sit elsewhere. Ask before signing up for a service, not after the first customer audit.

When the provider has a bad day

Large providers are generally more reliable than a company's own server room, but when one fails, thousands of customers wait together and nobody can send their own engineers. Leaders should decide how long each critical service can be down, whether it should run across more than one location, how the crisis team will communicate if email and chat are on the same provider, and whether the scenario has been rehearsed.

In the book, a regional outage of a little under five hours stops Meridian's customer app and retail ordering portal on a busy Saturday before a holiday weekend. App customers mostly reordered later; several smaller retailers went to a competitor. Running the portal in two regions would have added, illustratively, about fifteen percent to its hosting cost. The chief executive's lesson was that the earlier decision not to pay for this had been reasonable, but nobody had written down that it meant accepting a five-hour outage on the busiest weekend.

SaaS sprawl, shadow IT and setting up your part

Because cloud tools take minutes to start, companies accumulate far more of them than IT believes. Shadow IT, technology used without the knowledge of those responsible for it, is usually a sign of people trying to work, not misbehaving, and banning it simply hides it. Meridian expected about forty cloud services and found just over two hundred, including a free file-sharing account used for price lists, a survey tool holding thousands of customer email addresses and a former contractor who still had access. The book recommends a fast approval route, connecting tools to a central login so leavers lose access everywhere at once, finding services through network records, expense claims and existing tools, and announcing the search as an amnesty rather than an audit.

Finally, a well-configured cloud is usually safer than anything a mid-size company could run itself. The protections, however, are rarely on by default. The right question is not whether the cloud is safe but whether the company has set up its own part: multi-factor authentication for every account, automated configuration checks, logs that someone reads, tested backups kept outside the main account, and regular access reviews.

KEY POINTS

Key points

  • Ask for a one-page split of provider and company responsibilities, with a named owner for each of yours, for the five most important cloud services.
  • Know where customer and employee data is stored, including backups, and ask before signing up for new services.
  • Decide and record how long each critical service may be down, and fund multi-region resilience where it pays.
  • Keep a way for the crisis team to communicate that does not depend on the main cloud provider.
  • Connect cloud tools to a central login so access ends everywhere when someone leaves.
  • Confirm that multi-factor authentication, logging and automated configuration checks are on for every critical account.

QUESTIONS

Common questions

Who is responsible for security in the cloud, the provider or the customer?
Both, under what providers call the shared-responsibility model. The provider secures the cloud itself, such as data centres, hardware and core networks, while the customer secures what it puts in the cloud, including who has access, how data is configured and shared, and which security settings are switched on. Many serious cloud incidents happen on the customer's side of that line.
Is the cloud more secure than our own servers?
Usually, if it is set up well. Major providers have large security teams, see attacks across millions of customers and run heavily protected data centres that few companies could match. But protections such as encryption, logging, multi-factor authentication and backups often have to be switched on and configured by the customer, so the real question is whether your company has set up its own part properly.
Why does it matter where our cloud data is stored?
The location of data determines which country's laws apply to it and which authorities could request it. Data-protection laws such as GDPR and Japan's APPI place conditions on transferring personal data across borders, and many large customers ask where their data will be held. Check the main storage region, but also where backups, support access and new features process data.
What is shadow IT and should we ban it?
Shadow IT is technology staff use without the knowledge or approval of the people responsible for managing it, often free or credit-card cloud tools. Banning it rarely works because people who need to get work done find another route out of sight. A better approach is to make approved tools quick to obtain, connect services to a central login, and treat discovery as an amnesty that brings useful tools into the open.
How should we prepare for a major cloud provider outage?
Decide for each critical service how long the business can tolerate it being down, and pay for running it across more than one location where the business case is clear. Make sure the crisis team can communicate if email and chat run on the affected provider, and rehearse a day-long outage. Write down any outage risk you choose to accept so it can be revisited later.