IPMI, Bandwidth, and Remote Hands: Questions Before You Sign Dedicated
TL;DR: Spec sheets advertise cores and RAM; outages are won or lost on remote access, network fair-use rules, and how fast a human in the cage will reseat a NIC. Ask these operational questions while comparing dedicated hosting providers—before the contract auto-renews.
Out-of-band access is not optional
IPMI, iDRAC, or iLO lets you fix a hung kernel without a drive-across. Confirm:
- Whether out-of-band is included or a paid add-on.
- How you authenticate (VPN, jump host, separate credentials).
- Whether HTML5 consoles work from your laptop without proprietary plugins that break every quarter.
- Whether the management network is isolated from the public internet by default.
Test console access during the trial window. A feature that exists only in a PDF fails the night your SSH daemon will not spawn. Practice mounting a rescue ISO once so the first real emergency is not also a tutorial.
Remote hands SLAs in plain language
“Remote hands” can mean anything from “we reboot when ticketed” to “we will swap the drive you shipped.” Get response targets for business hours vs nights, fees per incident, and whether media destruction or loaner disks are available. If your app is revenue-critical, price a tighter SLA explicitly—do not assume the marketing badge covers midnight GPU reseats.
Ask who escalates when the first-line technician cannot identify your chassis. Mislabeled cages are rare until they are not.
Bandwidth: unmetered vs honest caps
Unmetered often means fair-use on a shared uplink. Ask for the commit speed, burst policy, and what happens if a software update CDN-sync spikes egress. Metered plans need clear overage rates. For video, large backups, or game downloads, model the 95th percentile of your current cloud egress and compare.
Also confirm IPv4 availability and pricing. A bargain server that charges heavily per extra IP can erase savings if you host many SSL sites without modern SNI-friendly setups (most stacks are fine—still verify). IPv6 readiness matters if your audience or partners already prefer it.
Cross-connects and private networking to adjacent racks can matter for hybrid storage. If you plan a second box later, ask about free or cheap private VLANs now.
Network neighborhood and DDoS posture
Dedicated does not automatically mean clean traffic. Ask whether basic DDoS filtering is included, what thresholds trigger null routes, and how you opt into scrubbing. If you run game servers or politically spicy content, assume you need an explicit mitigation conversation—not a forum rumor.
Traceroute from your users’ regions during evaluation. A cheap cage with a congested peering path feels slower than a slightly pricier host with better routes. Run the traceroute at peak evening hours, not only at 10 a.m. local datacenter time.
Hardware replacement and spare parts
Drives fail. Power supplies fail. Clarify replacement timelines, whether hot spares sit on site, and who initiates RMA. If you run software RAID, know the rebuild impact on production I/O. If the provider manages hardware RAID, know who gets alert emails when an array degrades.
Document serial numbers and photos of the front panel on day one. Inventory clarity speeds every future ticket.
Contract gotchas that show up month thirteen
Watch for early termination fees, automatic renewals, and upgrade paths that require full reinstalls. Ask how long they keep canceled disks before wipe, and whether you can pay for a witnessed destruction certificate if compliance requires it. Export a full backup to your own storage before any cancellation ticket.
Clarify jurisdiction and abuse desk contacts. A dedicated host that is slow to answer copyright or spam complaints can take your clean neighbors down with a shared uplink incident.
A pre-sign checklist you can paste into email
- IPMI/iLO included? Test account during trial?
- Remote hands scope, hours, and fees?
- Bandwidth policy with numbers, not adjectives?
- DDoS basics and null-route behavior?
- Disk replacement SLA and backup expectations (yours vs theirs)?
- Renewal terms and data wipe policy?
Trial week habits that reveal the truth
During any money-back or trial window, intentionally break things safely: drop the SSH service in a maintenance window and recover via IPMI; open a remote-hands ticket for a non-urgent reseat question and time the reply; push a large backup off-box and watch whether fair-use warnings appear. Providers that shine on sales calls sometimes stall on these boring drills—better to learn that before production DNS points at the chassis.
Save ticket transcripts and traceroute screenshots in the same folder as the signed quote. Renewal negotiations go better when you can show measured behavior, not only marketing PDFs.
Cores are easy to compare; operational reality is not. Put IPMI, bandwidth math, and remote hands in writing, then choose among shortlisted machines on Best Dedicated Web Hosting Server with eyes open about who helps when the OS will not boot.
