Managed DDoS testing is usually described as the hands-off option: a provider plans the attack simulation, runs the traffic and sends a report, and your team watches. That holds for the traffic, and for very little else, because the parts of a DDoS test that decide whether its result can be trusted stay with the organisation being tested whoever runs the generators. The useful question about a managed test is not who presses start. It is where the line between the provider's work and yours is drawn, and whether both sides drew it in the same place.
What the provider runs, and what stays with you
In a managed test the provider brings the traffic. It drafts the vectors and rate steps against your stated goals, operates the generating infrastructure from its own sources, runs the steps in the agreed window, stops on command and writes up what it sent.
Everything that gives the result meaning sits on your side:
- Permission. The asset owner signs, and so does every third party in the path. A provider can draft the paperwork; it cannot sign for your CDN or your ISP. The rules a cloud provider publishes apply to the test traffic whoever runs it, and authorized DDoS testing sets out what AWS, Azure and Google Cloud each require.
- Scope. Only you know which addresses carry revenue, which are shared with another business unit, and which sit behind a dependency nobody drew on the network diagram.
- The stop line. The abort criteria are yours to set, and someone on your side must be able to call stop and have it obeyed.
- The view from inside. The provider sees what it sent. Only your own telemetry shows what arrived, what your mitigation did with it, and what real users experienced.
- The fixes. No report closes a finding. Your change does, proved by a second run.
Managed versus self-serve
A self-serve platform moves the traffic to your side of the line as well. Your team picks the vectors, sets the rates and runs tests on its own calendar, and it carries the work of interpreting what happened.
Each suits a different situation, and neither is the better product in general. A managed test fits a large, infrequent exercise that needs several third parties coordinated, or a team that has never run one and wants the first done with people who have. Self-serve fits a team with the skill in house and a stack that changes often enough that waiting for a booked engagement leaves gaps. Many organisations use both: a managed test to establish the baseline, then their own repeats in between.
The delivery model says nothing about what the traffic does. A managed test can be a fixed script, and a self-serve platform can be sophisticated, so the methodology has to be judged on its own.
What to settle before a managed test
A few questions decide whether the engagement produces evidence or just an invoice.
Who calls stop, and how fast does it take effect? Name the person on your side, the channel, and the time from their call to the last generator going quiet. Then check that time during the test itself.
Will both views meet in the report? The provider's record of sent traffic and your record of what arrived should be reconciled step by step. A DDoS test report built only from the sending side says what was fired, not what it did.
Can it be repeated exactly? Ask for the parameters in a form you keep: vectors, rates, hold times, source spread and timing. Without them a DDoS retest after a fix compares two different tests.
Who tells the third parties? A CDN or carrier that sees an unannounced flood will treat it as one. Decide who makes those calls, and before which hour.
Do the sources look like an attack? Per-source rate limits and reputation filters are only really tested by traffic spread the way real attacks spread. A test from too few addresses measures those controls on an easy day.
Why the delivery model belongs second, after the question of what the traffic actually does, is argued in DDoS testing approaches compared, and it applies to a managed engagement as much as to anything a team runs itself.