Computing resource notification domain in network

Abstract

This document introduces the definiation and the requirements of notification domain, and also introduces the process of service scheduling decision based on the notification domain in the network.¶

1. Introduction

With the new services arising, the computing and network convergence has become a new trend. The network could schedule the service traffic considering the computing metric.There are several ways to get the computing information. By the computing information notification schemes, the routers in the network can synchronize their connected local computing information among themselves, eventually forming a global complete computing information. There may be multiple computing information notification schemes, one is that computing information status notification could made by extending BGP (Border Gateway Protocol) to the whole network, without defining the scope of the notification, so that all the routers on the whole network need to maintain the computing information of the whole network, and the routing table is large, which causes the burden to the network and routers. Exending BGP without limiting the scope of computing information notification increases the amount of computing information notification in the network, while the service traffic steering is always among near area. So to reduce the useless computing information notification in the network, and to improve the effciency of the computing information notification scheme is important.¶

This document introduces the definiation and the requirements of notification domain, and also introduces the process of service scheduling decision based on the notification domain in the network.¶

2. Definition and Requirements of the notification domain

The routing system contains the control plane and the forwarding plane. To achieve the computing and network coordination, the control plane needs to obtain the computing information of the servers for a service, and monitor the tunnel/ policy statuses to the potential Egresses for the service. As described in [I-D.yao-cats-ps-usecases], the representation and encoding of computing metric is crucial, which is conveyed to routing system to support the components to act upon. The routers send the computing information to the routers located in the given notification domain. If router 1 is outside of the notification domain of router 2, then router 2 will not send computing information to router 1. The notification domain needs to be defined accurately, the requirements contains several aspects as described below. o Should support to determine the notification domain based on the computing service topology. The computing service topology contains information about the identity of the computing service, as well as information about the location where the computing service is deployed .o Should support to determine the notification domain based on geographical location, and/or the division of areas with the same service deployment into a notification domain based on service dimension. o Should support to determine the notification domain according to the geographical location and then the notification relationship between each autonomous domain is determined according to the existing AS domain in the IP protocol. o Should support that when the size of the notification domain is determined, according to the network status information, the notification domain is expanded when the network status is good, and the notification domain is reduced when the network status is bad. o Should support that when the notification domain is divided according to services, the size of the notification domain is determined according to the number of nodes deployed in the notification domain.¶

3. Process of service scheduling based on the notification domain

The computing information notification method is proposed to reduce the amount of computing notification information.¶

Step 1: The central controller determines the notification domain¶

Option 1: The central controller could determine the notification domain based on the computing service topology. The generation of computing service topology includes the identification information of computing service and the location information of computing service deployment, and advertising the service topology information; and then advertises service status information on the established service topology. The notification domain could be defined based on the division of areas with the same service deployment into a notification domain based on service dimension. During implementation, the computing service topology is generated based on the network topology and computing node topology generated by BGP-LS; and the service topology information is advertised through BGP-LS and DHCP by adding computing service information to the node attribute field.¶

Option 2: The notification domain could be defined based on the division of adjacent domains into a notification domain based on geographical location, and the central controller obtains the AS autonomous domain division strategy of the whole network and determines the size of the notification domain by combining the network status information of different regions. The network status information is reported by nodes to the central controller.¶

Option 3: The central controller receives the inter-domain router network state information, service deployment information, as well as generating capacity after list, according to the different domain partition strategy to subscribe to the node network status information or services generated after the deployment information, and distributed to the inter-domain router for its configuration.¶

Step 2: The central controller advertises the notification domain partitioning policy. The notification domain partitioning policy carried through Netconf and Yang management protocol is delivered. In practice, the notification domain is identified as follows:¶

o The number of AS through which the computational update information passes is set, and the hop limit is set to identify it. Or,¶

o Routers in the same advertisement domain are divided into the same community by using the BGP community attribute to identify the advertisement domain¶

Step 3: The central controller advertises computing information to nodes in the notification domain.¶

Step 4: When the central controller schedules service requests, it selects the computing nodes that need to obtain real-time service status information through the generated computing service topology and sends the detection packet of service status information.¶

Step 5: The notification domain adjustment policy generation When adjusting the size of the notification field, it further includes: The notification domain adjustment policy carried through Netconf and Yang management protocol is delivered based on one or a combination of the network status information, service scheduling result feedback information, and service deployment. When the size of the notification domain is determined, according to the network status information, the notification domain is expanded when the network status is good, and the notification domain is reduced when the network status is bad. When the notification domain is divided according to services, the size of the notification domain is determined according to the number of nodes deployed in the notification domain.¶

7. Normative References

[RFC2119]

Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997, <https://www.rfc-editor.org/info/rfc2119>.