Load sharing in Call Server clusters

Дослідження Роботи Одиночного Сервера У Кластері

“…Цей показник визначається під час відгуку системи. З точки зору вхідного трафіку для кластера серверів найкращим індикатором є затримка 95-го процентілю -95-й POD (percentile origination delay) -POD P , яка використовується для виявлення перевантаження [12].…”

Load Sharing in a Distributed IMS Architecture

“…Load balancing mechanisms distribute incoming calls across the servers of one site to achieve equal load distribution and high availability [2], [7]. In such scenarios, load originating from several access networks concentrates on one server cluster leading to statistical gain.…”

“…In [2] such concepts are transferred to centralized call server clusters, however with static distribution schemes and a small number of call servers only. The IMX scenario is different from web and call server scenarios, since IMX nodes will always stay in the path for processing or policing tasks.…”

“…The two basic steps of load information gathering and resource allocation can be further subdivided in a measurement and update policy as well as a transfer, location and redirection policy (see also [2]). The measurement policy defines how often the system performs load measurements, their granularity and aggregation strategy of values (mean, peak,...) resulting in a node-local load view.…”

“…Load sharing is not a new concept and is already widely used with PSTN equipment (e.g., [2]) and large-scale server clusters [3]. Local load balancing schemes for IMS servers have also been studied [4], and there are ongoing standardization activities concerning overload signaling [5].…”