I think maybe an example would be easiest. Let's say the client asked 'can your system respond within 10ms to requests?'. I'd say 'no it can't guarantee this'.
I'd then discuss how the system architecture couldn't give any hard latency guarantees, but that depending on the load and hardware they purchased we would generate different latency characteristics, say, 98% of requests processed within 10ms.
This would lead to an offline discussion and sharing some profiles for existing hardware configuration we'd tested, and maybe some estimates of the hardware costs to meet their requirements.
Comments
I think maybe an example would be easiest. Let's say the client asked 'can your system respond within 10ms to requests?'. I'd say 'no it can't guarantee this'.
I'd then discuss how the system architecture couldn't give any hard latency guarantees, but that depending on the load and hardware they purchased we would generate different latency characteristics, say, 98% of requests processed within 10ms.
This would lead to an offline discussion and sharing some profiles for existing hardware configuration we'd tested, and maybe some estimates of the hardware costs to meet their requirements.