Dedicated server specifications are easy to compare and tell you little about the experience of running one. What matters is what happens when something fails — and that is almost never on the spec sheet.
Out-of-band access
When the operating system will not boot, SSH is useless. You need console access independent of the OS — a remote management interface that lets you see boot messages, enter BIOS and mount installation media.
Without it, every problem requiring a look at the console becomes a support ticket and a wait. With it, most are ten minutes of your own time.
Ask whether it is included or chargeable, and whether access is self-service or on request.
Hardware failure: what is committed
Disks fail, memory develops faults, power supplies die. The question is what happens then.
- Is there a stated replacement time, and what is it?
- Are spares held on site, or ordered when needed?
- Who reinstalls and restores if a disk fails?
- Is there RAID, at what level, and is it monitored — a degraded array nobody notices is a failure waiting for its second disk
Monitored RAID with a stated replacement commitment is worth more than a faster processor.
Bandwidth terms, read carefully
"Unmetered" and "unlimited" usually mean a port speed with a fair use policy. Ask:
- Port speed, and whether it is shared or dedicated
- Whether there is a transfer allowance and what exceeding it costs
- Whether inbound and outbound are counted separately
- Whether traffic is throttled or charged at a threshold
For a server serving media or backups, this matters more than the hardware.
DDoS: what is actually included
Most providers advertise DDoS protection. The differences:
- What volume is absorbed before it becomes a problem
- Whether mitigation is automatic or requires a ticket
- Whether your IP is null-routed during a large attack — which stops the attack and also stops your service, and is a common policy worth knowing about in advance
- Whether mitigation beyond a threshold costs extra
Null-routing policy is the important one. Discovering it during an attack is a poor time.
Where the machine physically is
Latency to your users depends on it, and so does any requirement about data location.
Ask for the data centre location and its facility standards — redundant power and cooling, network carrier diversity. A dedicated server in a facility with a single power feed is not a resilient server.
For an Indian business serving Indian users, a server in India is usually meaningfully faster than one abroad. That is measurable — test it before assuming either way.
Backups are usually not included
Dedicated servers typically come with no backup at all. The hardware is yours, and so is the responsibility.
Ask whether backup space is available, where it physically sits relative to the server, and whether it is included. Backup on the same machine, or in the same rack, protects against less than it appears to.
Support scope, precisely
Where does the provider's responsibility end? Typically at the hardware and network, with the operating system and everything above it yours. Some offer managed tiers that extend further.
Establish this clearly before an incident, because the worst moment to discover the boundary is while trying to get something fixed.
About the author. Kislay Chandan has worked with schools since 2014. Rather than gathering requirements from the outside, he worked inside a school to see what actually goes wrong day to day, and built what he found into the software. He still works with Shemrock Glitters in Patna — most changes to the product start with something that came up there.
Gallery
