How to evaluate service caching？
Shenzhen 10Gigabit Ethernet Technology Co.,ltd
Empowering Communication, Securing Networks - Your Trusted Partner for a Connected World
Fancy Wang 0702 2023
Evaluation method of service cache
The service objects of edge computing are end users, and will focus on edge network resource utilization, edge system stability and edge system reliability in mobile scenarios.
In edge network service caching, from the perspective of operators, the first issue that needs to be considered is how to increase the number of service edge users under the condition of fixed resource costs. Due to the differences in the needs of edge users, the glasses packet core network (EPC) has gradually become unable to handle the service needs of different users.
Obviously, resource utilization is the key to meeting user needs and increasing the number of users, which has also promoted a large number of related research on on-demand caching/deployment of network functions.
Each network slice consists of a set of independent virtual resources with their independent topologies, and user classes are divided by their request traffic and their required network functions. Each network slice can act as a logical "private network" for service customization, and all network slices share the underlying basic physical resources.
Different scenarios have different requirements in terms of latency, bandwidth, and storage. Using network function virtualization technology can separate network functions from dedicated underlying hardware resources, and slices are not restricted by fixed locations. Therefore, the network slice of the required service can be flexibly deployed according to the demand, so as to meet the hydrangea of the specific service and further improve the quality of service.
Although some progress has been made in the allocation of network function resources in the cloud-centric network, how to provide services for more users with limited distributed resources in the edge network is still an urgent problem for edge network operators to solve.
In network slicing, although services are logically divided, different slices still share the underlying hardware resources. In order to save energy or reduce communication waiting time, it is necessary to merge VNFs into the same server, which is called VNF merging .
Research at this stage shows that VNF merging will lead to a decline in throughput and timeliness. Because some 5G network slices have very strict performance requirements, it is easy to cause mutual interference between VNFs.
Although the edge cloud has greatly enhanced the computing power of mobile users, the uncertainty of edge user movement has brought new challenges to service deployment. Since the edge server is very close to the user, the impact of user mobility on service delay is more obvious than its impact on cloud computing.
领英推荐
Therefore, in order to optimize user experience, edge computing operators need to do much more than just push the service functions in cloud computing to the edge of the network. In the edge network, in order to ensure the service continuity when users cross different edge service networks, an effective mobility management solution should be adopted at the edge of the network to ensure the stability of the edge system:
A: The stability of user services. To ensure the uninterrupted service of users during the server switching process.
B: Stability of the edge network: Minimize the overhead of service migration while meeting the needs of mobile users. In order to improve the experience of edge network users, the network services required for mobility can be dynamically redeployed between edges to meet user mobility.
But dynamically redeploying web services is not easy. On the one hand, user-perceived delay in dynamic service deployment is jointly determined by communication delay, computing delay, and service migration time. If each user’s required service is deployed on the nearest edge server node, it will result in:
a: Some edge nodes may be overloaded, resulting in increased computation latency.
b: Frequent redeployment of services will increase the service migration time and increase the user's end-to-end delay.
c: Migration of services will cause interruption of user services and reduce user service experience.
On the other hand, following user mobility requires frequent service migration among multiple edge servers, and such frequent service migration will incur additional operating costs. Therefore, an effective caching strategy for dynamic services should be cautious:
1) Comprehensively consider communication delay, computing delay and service migration delay to minimize user delay.
2) Trade-off between performance and cost in a cost-effective manner to increase system stability.
The reliability of the marginal system is another important index for the evaluation of the marginal system. As mentioned above, considering the limited edge network resources and the diversity of terminal devices, operators generally use network function virtualization technology to reduce the impact of dedicated hardware and improve the elasticity and flexibility of edge server resources. Despite these advantages, the separation of network function software from the underlying dedicated hardware poses several challenges to system reliability.
Specifically, in most existing VNF systems, a VNF runs as an instance on a virtual machine, and its resources are controlled by the underlying hypervisor, so any failure of the hypervisor will cause the VNF running on it to unavailable. When multiple VNFs are connected together to provide overall network services in the form of a service chain, any VNF failure in this service chain will make the entire service unavailable. Therefore, it is necessary to add VNF backup in the edge network to improve the effectiveness of the service chain,