Guaranteed-Rate (GR) servers, such as Weighted Fair Queueing (WFQ) and its variants, have been widely used to give mainly bandwidth guarantees and consequently delay guarantees for real-time flows provided that their arrivals are upper-bounded. Problems may arise if a bursty traffic with a small service share needs a specific short delay. In fact, the higher the service share is, the lower the delay the flow gets. However, WFQ and its variants are share-driven servers and no temporal constraint is considered in the scheduling process. Therefore, having in mind that real-time streams could tolerate some deadline misses according to their (m,k)-firm constraints, we propose a new scheduling technique called (m,k)-WFQ that extends WFQ to also consider (m,k)-firm temporal requirement. Analytic expressions using Network Calculus theory are derived to give deterministic upper bound on delay provided by the (m,k)-WFQ scheduler. Theoretical Results and simulations show lower average and maximum delays provided by the proposed scheduling algorithm, without much degrading bandwidth fairness.