“DDoS protected” appears on nearly every dedicated server sales page, and almost none of them quantify it. The meaningful differences between providers are not whether they mitigate — most do — but how much scrubbing capacity stands behind the mitigation, what baseline traffic is always-on versus on-demand, and what the SLA actually promises when an attack exceeds capacity. Those three variables decide whether an attack costs you ten minutes or ten hours.
This breakdown separates the marketing from the engineering so you can ask three precise questions on a sales call and get an answer that distinguishes a real mitigation stack from a null-routed afterthought.
Scrubbing Capacity: The Number That Is Never on the Page
Scrubbing capacity is measured in Gbps of clean traffic the mitigation cluster can absorb while filtering. Tier-1 providers publish cluster totals of 1–15 Tbps, but the relevant figure is what is provisioned for your service tier, not the global cluster. A provider with 10 Tbps total may give you a 2 Gbps guarantee and a 10 Gbps burst ceiling.
Ask for both numbers in writing. The guarantee determines what is defended unconditionally; the burst ceiling determines what you get for the first few minutes of an oversized attack. Attacks under 100 Gbps are routine now, so any dedicated tier advertising a guarantee below 10 Gbps is relying on upstream carrier filtering you do not control.
Scrubbing capacity is also shared. Your guarantee is a reservation against a cluster that serves every customer on the node, so a simultaneous attack on three tenants can leave you with less burst headroom than the number on your contract implies. Ask whether guarantees are backed by reserved capacity or by best-effort aggregate capacity — the answer changes how much protection you actually have during a large campaign.
Always-On Baseline vs On-Demand Detection
Baseline protection means traffic passes through scrubbing continuously and clean traffic reaches you with a small added latency — typically 1–5 ms. On-demand protection routes traffic normally until an anomaly triggers a BGP diversion, which adds 30–180 seconds of detection plus a transient where inbound traffic is misdirected.
For game servers and anything with persistent UDP sessions, always-on baseline is worth the latency. On-demand detection misclassifies legitimate traffic spikes — a patch-day login surge looks a lot like a SYN flood to a threshold-based detector. If you run game workloads, check the dedicated hosting configurations built for game and realtime traffic rather than accepting a generic on-demand tier.
What the Scrubbing Stack Actually Filters
A modern scrubbing stack handles four classes: volumetric floods (UDP amplification, DNS reflection), protocol attacks (SYN, ACK, fragment), application-layer floods (HTTP request floods, slowloris), and state-exhaustion attacks against connection tables. Coverage of the first two is table stakes; coverage of the last two is where providers diverge.
Application-layer filtering requires the mitigation layer to understand your protocol. A provider that only does L3/L4 will pass an HTTP flood straight through because each request looks legitimate at the packet level. Ask specifically whether they do L7 inspection and whether it is included or an upsell.
The scrubbing route also determines your egress bill. Some providers meter clean traffic returning from the scrubbing centre, which means an attack that generates 50 Gbps of inbound junk can still cost you in outbound leg-back charges. Confirm the billing treatment of scrubbed return traffic before you assume mitigation is free, and factor it into the dedicated server plans with included DDoS filtering you are comparing.
Clean Traffic Latency and the Route Back
Scrubbing introduces two latency costs: the detour to the scrubbing centre and the inspection overhead. If your scrubbing centre is in Frankfurt and your server is in Amsterdam, clean traffic detours several hundred kilometres. Providers that run scrubbing inline at the edge avoid this; providers that centralise do not.
Measure it yourself before committing. Ping your allocated IP from three regions and compare against a provider without mitigation in the same city. A 2 ms penalty is invisible; a 25 ms penalty will break competitive game servers and hurt database replication.
SLA Language: What Is Actually Guaranteed
Read the SLA clause on mitigation carefully. Three distinct promises hide under “DDoS protection SLA”: uptime of the mitigation infrastructure, time-to-mitigate once an attack is detected, and financial credit for failed mitigation. Many SLAs guarantee only the first and remain silent on the second and third.
A defensible SLA names a detection window (usually under 60 seconds), a mitigation window (under 5 minutes for volumetric), and a credit schedule. If the clause does not mention attack response time, assume there is none and price the risk accordingly.
When the Attack Exceeds Capacity
Every mitigation stack has a ceiling, and the behaviour above it matters more than the ceiling itself. Good providers shift to a blackhole route for the specific attacked IP with a short timeout and notify you; poor providers null-route the whole /24 or let the link saturate and take down neighbouring customers.
Ask what happens at 120% of guaranteed capacity. If the answer is “we null-route your IP for 30 minutes and email you”, that is acceptable and standard. If the answer is vague, the provider has not thought about it — which means neither has their on-call engineer at 3 a.m.
Questions to Ask Before You Sign
Five questions separate a real mitigation tier from a marketing claim: What is my guaranteed scrubbing capacity in Gbps? Is baseline filtering always-on or triggered? Do you filter at L7? What is the mitigation time window in the SLA? What happens above guaranteed capacity? Write the answers into your order confirmation.
If a provider cannot answer all five in one email, the mitigation is likely resold from an upstream carrier with no service-level commitment behind it. That is not automatically disqualifying for a low-risk blog, but it is for anything generating revenue per minute of uptime.




Leave a Reply
You must be logged in to post a comment.