在網站閘道器上提供差別服務品質之多重資源請求排程

Abstract

差別服務品質是網站經營者提供給客戶不同層級服務的一種方式，傳統HTTP請求排程的方法可以達成這個目標，但是它們排程請求只能管理一種伺服器資源，例如CPU或Disk I/O，但實際上伺服器在處理一個請求時會消耗多種資源，單一資源的排程法會導致資源的浪費或系統的過載。本論文提出了一個名為mQoS的多重資源請求排程演算法來將伺服器上的多種資源使用量作差別管理，mQoS排程演算法是部署在網路閘道器上，因此它對客戶端及伺服器端是透明的。mQoS排程器中是由數個子排程器及一個主排程器所組成，每一個子排程器管理一種伺服器資源以便在各個服務類別之間差別資源使用量，主排程器則檢查每種伺服器資源的可使用量並觸發適當的子排程器以平衡各伺服器資源的使用。mQoS排程演算法的設計想法是源自於傳統的Deficit Round Robin封包排程演算法，每個服務類別都對應到一個deficit counter以紀錄其未使用的伺服器資源量，而子排程器中的deficit counter可以被任何一個子排程器減值，主要原因為請求是消耗多重資源而非單一資源。mQoS閘道器的實作是架於Squid及Linux之上。在性能評估方面，將比較mQoS、無排程(nQoS)及單一資源排程(sQoS)以展現mQoS排程的效果。在以差別服務比例6:3:1分配給三個服務類別的例子中，mQoS排程準確地分配各種伺服器資源，另外，mQoS排程的總伺服器吞吐量比sQoS排程增進了21%。mQoS的平均使用者察覺延遲比其他兩種方式較短。

Differentiated quality of service is a way for a Website operator to provide different service levels to its clients. Traditional HTTP request scheduling schemes can achieve this, but they schedule requests to manage only one server resource, such as CPU or Disk I/O. Actually, processing a request on the server will consume multiple resources. In this paper, a multiple-resource request scheduling algorithm, called mQoS, for differentiating the utilization of the server resources is presented. The mQoS scheduler consists of several sub-schedulers and a main scheduler. Each sub-scheduler manages a server resource to differentiate the utilization among the classes. The main scheduler checks the availability of every server resource and triggers an appropriate sub-scheduler to balance the utilization of server resources. The idea of the mQoS scheduling algorithm is derived from the traditional deficit round-robin method. There are some deficit counters in a sub-scheduler. However, a deficit counter of a sub-scheduler can be decremented by other sub-schedulers because a request would consume multiple resources. The implementation of the mQoS gateway is based on the Squid and Linux. In the evaluation, the mQoS scheduling is compared with no scheduling (nQoS) and single-resource request scheduling (sQoS). The mQoS scheduling reveals the accurate differentiation on every server resource. In addition, the total server throughput in the mQoS scheme is improved by 21%, compared with the sQoS s. The average user-perceived latency of the mQoS scheduling is shorter than other schemes.