Write-ups from deployments we have actually run, organised by section. Specific problems, the reasoning behind the approach, and what it cost to get wrong. Client names are withheld under confidentiality.
Page 5 of 8
Prometheus is the default metrics store for good reasons, and it has a documented limitation businesses run into by using it for the wrong job. Knowing what it is not for is more useful than another installation guide.
Most server monitoring watches CPU, memory and disk, which are the three things least likely to be the actual problem. The signals that predict outages are duller and are usually not collected at all.
The distribution argument is the least important part of this decision. What matters is the support lifecycle you are committing to, whether anyone here can administer it, and whether the machine will still be patched in year four.
Virtualisation is treated as settled and mostly is. What is not settled is what it actually provides — because several things routinely credited to it, including availability and backup, it does not provide at all.
Almost every growing business is already hybrid, usually by accident. Making it deliberate means answering one question per workload — and the answer turns on data gravity, latency and exit cost far more than on any cloud-versus-on-premises argument.
Enterprise network design usually arrives as a three-tier diagram from a vendor guide, which is the wrong shape for almost every growing business. The decisions that matter are the addressing plan, where the failure domains sit, and how much uplink you actually have.
Almost every business has antivirus and believes endpoints are handled. The gap between having it and being protected by it is where most intrusions live — in the machines that stopped reporting, and in the attacks that never involve a malicious file at all.
Audit, assessment, scan and penetration test are used interchangeably and are four different things with four different prices. Buying the wrong one is the most common way businesses spend a security budget and learn nothing they did not already know.
Switches and routers are the most privileged and least maintained devices in most estates. Cisco's own hardening guidance divides a device into three planes, and that division turns an unbounded task into three short, checkable lists.
Most Windows Server advice is a list of settings. The things that cause real outages are lifecycle dates nobody tracked, roles piled onto one machine, and a domain controller nobody has ever restored. Those are the ones worth the attention.
This stopped being a technical comparison some years ago. For a growing business both hypervisors do what is needed, and the decision now turns on licensing, on the skills you already have, and on how much a migration would cost if you chose wrong.
This is normally compared as a feature list, which is why it is normally decided badly. The two platforms have different licensing shapes and different maintenance obligations, and those two things — not the feature grid — determine which one is cheaper for you over five years.
We typically respond within 4–8 business hours.
