Bill Rubino/ACG Research
01 Jun 2010
2. Scaling for simultaneous EPC sessions. Sessions refer to the number of simultaneous sessions that can be established per EPC. A service provider can have the best EPC with great performance and signaling capacity, but if it cannot scale to a large number of simultaneous sessions, throughput and signaling are pointless because the EPC will be dropping sessions if it is over capacity and reaches the ceiling for the maximum number of simultaneous calls. As with signaling, headroom built into the session capability of the Evolved Packet Core will address this issue.
3. Throughput while using Deep Packet Inspection. The EPC must support the performance needs and the high bandwidth requirements for 4G -- throughput of 30Gbps or greater. Throughput while using Deep Packet Inspection (DPI) is another crucial element to address. When a carrier uses DPI, the question is, what is the network performance hit when it is enabled? DPI is a processor intensive application because all packets that flow through the EPC have to be inspected right up the OSI layer stack. This will affect overall throughput. Taking a serious performance hit of 40% or greater when DPI is enabled could cause problems for an operator.
4. Architecting the Evolved Packet Core for redundancy. All carriers want redundancy built into their networks. The concept of five 9s (99.999%) reliability originated from the requirements of the old analog networks and is still relevant for today's carrier networks. The evolved packet core must have some level of redundancy to provide reliability, however. Carriers need an EPC architected with a redundancy scheme that isn't going to break the bank in terms of deployment. Having a primary and secondary EPC device where the secondary device sits dormant until the primary fails is an expensive deployment scenario. EPC devices that have 1:N redundancy will help because a carrier will not need dormant secondary devices or have dormant secondary slots or blades that take up space within a chassis as with 1:1 redundancy. The goal is to provide reliability that is five 9s, not nine 5s, without additional equipment expense while doing it.
Bill Rubino is a principal analyst at ACG Research
This article originally appeared on SearchTelecom.com