Autonomic Networking in mobile wireless backhaulJuniper Networksdeanb@juniper.net
Operations and Management
UCAN BOFMobile backhaul networks that utilize microwave technology in transport are suspicious to seasonal and/or meteorological changes. For those reasons throughput can vary significantly. This draft provides problem statement and how autonomic networking can be applied to the problem.Microwave technology is one of the workhorses in MBH networks today. Unlicensed microwave links can be set up in days rather than the weeks or months it might take to implement additional wireline capacity for backhaul. Even licensed links, while requiring some mildly time-consuming bureaucratic approval, still easily outpace the time-to-deployment of wireline alternatives. Fiber may offer unlimited bandwidth, but the tradeoff is in time and cost savings. Microwave’s improvements in bandwidth, capacity, and reliability in the past few years have made it an ideal interim broadband connectivity solution during the often lengthy process of deploying fiber. In fact, these improvements go as far as to establish microwave as a legitimate permanent alternative to fiber. Although its many benefits, because of other restrictions that microwave links have, they can't be utilized at maximum. OSPF/MPLS networks that are overlayed on top of microwave transport and provide additional benefits of packet routing and switching to mobile backhaul.
MBH: Mobile BackhaulBTS: Base Transceiver StationOSPF: Open Shortest Path FirstMPLS: Multi Protocol Label SwitchingPLMN: Public Land Mobile NetworksCoS: Class of ServiceMTTR: Mean Time To RecoveryMNO: Mobile Network OperatorIDU: In Door UnitODU: Out Door UnitSNMP: Simple Network Management ProtocolIP: Internet ProtocolIPv4: Internet Protocol version 4IPv6: Internet Protocol version 6Mobile backhaul (MBH) networks utilize microwave networks to transport traffic back to Public Land Mobile Networks (PLMN). Base transceiver stations (BTS) and/or eNodeBs connect to a device that has multiple microwave connections to PLMN. Not each link has the same throughtput and the quality of the link varies with different factors,like air temperature, percipitation, vegetation, etc. Today those networks are overlayed with OSPF/MPLS networks and although OSPF provides automatic redirection of traffic in case of primary link failure, it reduces network throughput, as microwave link bandwidth slowly degrades, due to rain, snow, tower bending due to wind and/or temperature, vegatation growing. During the link degradation period, the throughput of MBH part is going down and the overall service is impacted. Being able to detect the degradation of the microwave link bandwidth and redirect traffic over higher throughput links is very beneficial to mobile network operators.
As MBH links are lowering the bandwidth, the user experience is impacted, as the data hungry applications are not served with usual quality of service and latency is increasing, due to dropped packets. MBH network administrator are not getting real time picture (usually today they see average link performance over 15 minutes period) and with users being highly mobile, they can't react to the challenges in the network. Administrators should be able to set intended policy on device, based on which device wwould start changing network forwarding parameters based on which the current traffic would be routed via links with moste throughput. With monitoring the link statistics, device can change forwarding and routing parameters in realtime based on the intended policy pushed on the device, without the need to interact with centralized management system, which would act based on sent link performance indicators. Such a network would improve end user experience, as well network administrators would be able to create better performing networks.Numerous parameters are involved in monitoring MBH, from microwave link performance, to miscellaneous OSFP and MPLS parameters. MNO has to look at KPI that will relate to those that may impact revenue negatively, such as unavailability and MTTR. One thing to note here is that much emphasis is usually placed on availability, while most times not enough emphasis is placed on reliability. In defining Key Performance Indicators effectively, KPIs must align with BTS availability information for a mobile operator.Microwave transmissionavailabilitycapacitydelayjitterAll OSPF interfaces have a cost, which is a routing metric that is used in the link-state calculation. Routes with lower total path metrics are preferred to those with higher path metrics. OSPF assigns a default cost metric of 1 to reference bandwidth and default cost metric of 0 to the loopback interface (lo0). No bandwidth is associated with the loopback interface. So if reference bandwidth is set to 1Gbps, it means that all interfaces faster than 1Gbps have the same default cost metric of 1. If multiple equal-cost paths exist between a source and destination address, OSPF routes packets along each path alternately, in round-robin fashion.Having the same default metric might not be a problem if all of the interfaces are running at the same speed. In MBH, microwave units will connect via ethernet to ethernet ports on routers and each link will have the same metric. That would not be a problem if all microwave links would have same performance, but links operate at different speeds, and it is very probable that traffic is not routed over the fastest interface because OSPF equally routes packets across the different interfaces.The autonomic agent agents collects operational statistics from ethernet ports to which microwave IDU is connected, as well from microwave ODU (local and remote) using SNMP. By collecting this statistics, optimal MBH OSPF agent can callculate links with best performance and change the metric value for each link accordingly.The section describers what information is needed to be provided by external entity, so that devices can operate autonomicly. The policy would have to set:reference bandwidth - example 1Gbpslow water mark threshold, at which point to change the metricIP address or addresses of local IDUIP address or addresses of local ODUIP address or addresses of remote IDUIP address or addresses of remote ODUwhich ethernet port is connect to which microwave linkThis list is not extensive and with more research it can be augmented with new parameters. The experience will show what parameters are important. There might be a need to put time restrictions between metric updates on the device or how often are statistics collected, as it is important not to negatively effect device forwarding capabilities.The device would interact with microwave IDU and ODU. It would interact with them via SNMP or some other protocol that allows to collect operating statistics of the microwave link. By collecting those statistics, it can compute the link perfomance. It is also possible to communicate with other autonomic device in MBH and to exchange information, so that devices can learn the whole topology in segment and performance of the microwave links in possible path. In Figure 2, a small example is shown how MBH router 1 is connected via microwave links to router MBH 2 and MBH 3. Microwave IDUs are connected via ethernet to MBH routers and each IDU has two ODUs connected. Microwave links usually have two beams in a link. Microwave IDUs send each incoming packet from MBH router 1 to each ODU connected to them, essentially copying packets over each beam in the microwave links. The terminating IDU C and D, on the other side, compare incoming packets from each ODU and drop the duplicate packets prior to forwarding the packet to their connected MBH routers. This mechanism allows collecting good operating statistics of the links, so autonomic agent on MBH routers can calculate end to end performance of each link, like latency, throughput, jitter, delay etc. This allows building performance topology on the MBH router by autonomic agent
The autonomic user agent should provide feedback data to centralized management system, so that new improved intent policies can be created. Most of the data doesn't have to be provided in real time, except for cases when microwave link failure happens and causes loss of data. This means that the autonomic agent didn't provide alternative path in time or all microwave links from the MBH router are down. Historical data, like what were the network conditions under which the autonomic agent enforcing the intent policies are very valuable, as well the performance topology from each device, as it allows to create whole performance view of the MBH.There are some vendors (NSN, Ericsson) that are trying to create self organizing networks, but the inteligence is always centralized, which prevents distribution of the network inteligence and using it for autonomic use cases.As this stage, author of the draft didn't look into security considerations of the use case.This document requests no action by IANA.