This is a creation in Article, where the information may have evolved or changed.
"Editor's note" The containerized ICT convergence platform that will be shared is a new cloud computing PAAs platform for future ICT systems, which is based on the lightweight virtualization technology of containers and an automated "microservices" management architecture that can effectively support the application's fast online and auto-scaling capacity, Maximize IT infrastructure resource utilization and reduce total cost of ownership (TCO). The future of the network is to the IT, cloud-oriented development, container and micro-service technology, perfect fit "network as a service", network slicing and other development concepts, will help to achieve more flexible, intelligent, efficient and open 5G new network.
In the 2015, NFV and containers were undoubtedly the hottest technologies, and were considered by many in the industry to be the future trend; in 2016, when NFV meets the container, what new spark will it collide with? What is the new experience when the containerized ct/it business is deployed within the same platform? Let's share with you, from 2015 to 2016, we did the work of containerized ct-it Fusion. It I don't explain, CT mainly refers to the telecommunications business.
The containerized ct-it Fusion system is a new cloud computing PAAs platform for future ICT systems, which is based on containers, a lightweight virtualization technology and an automated microservices management architecture that can effectively support application fast online and auto-scaling, maximizing IT infrastructure resource utilization and reducing overall TCO. The future of the network is to the IT, cloud-oriented development, container and micro-service technology, perfect fit "network as a service", network slicing and other development concepts, will help to achieve more flexible, intelligent, efficient and open 5G new network.
Here are some of the things we do in containerized ct-it Fusion from 2015 to 2016:
- 2015, we first put forward the containerized ICT integration solution
- Early 2016, we completed the containerized ICT Fusion prototype system
- April 8, 2016 to 9th, China Mobile held 2016 annual technical work, containerized ICT Fusion prototype system debut technology work will show.
- June 20, 2016 23rd, OPNFV Summit was held in Berlin, Germany. In the OPNFV POC Theater, China Mobile and red Hat jointly showcased the container-based ICT convergence platform
- June 29, 2016, the GSMA hosted the 2016 World Mobile Conference-Shanghai (Mobile Worlds Congress, called MWC) at the Shanghai New International Expo Center, at the 5G Exhibition of China Mobile Research Institute, we demonstrated the container-oriented ICT fusion prototype system.
The idea of first producing a container for the network was derived from the 2015, we are conducting NFV IMS testing, and we are also conducting technical research and validation work on containers and DCOs, and on the application scenario, the most suitable container-based application is undoubtedly the so-called cloud-native application, or MicroServices, whose main feature is distributed , no single point of failure, no state, this and we test some of the manufacturers of IMS network elements are so consistent, and the container of lightweight, high-performance, fast start and other advantages, also happens to be required by the network element, a few minutes to deploy a set of core network elements may really be from the dream into reality. As a result, a new combination was born, NFV CONTAINER=NFC.
Theory + practice can create the truth, on the basis of theoretical analysis, feasibility verification is essential, and surprises are always everywhere, try to innovate the road is not always so lonely, open source IMS Project Clearwater is also doing similar work with us. Clearwater is an open source implementation of IMS network elements, whose main goal is to develop a set of IMS network elements that are suitable for deployment on the cloud, such as AWS, to provide voice, video, and messaging services. Clearwater is architected to design individual components in a microservices manner and provides container mirroring in the Docker format, which we ported to the Kubernetes platform with load balancing, disaster recovery, and support for resource monitoring to make it more suitable for service
Test environment
Hardware
Red Hat 3D Print Demo system with 3 servers.
Software
- Operating system: Red Hat Enterprise Linux Server Release 7.2 (MAIPO)
- Container Engine: docker-1.8.2-10.el7.x86_64
- Container layout: kubenetesv1.1.0-origin-1107-g4c8e6f4
Clearwater Architecture
Clearwater Architecture is divided into six major components, Bono, Ralf, Sprout, Homer, Homestead, Ellis, as shown, Sprout as the SIP Router, is the core business processing module, is also the increase in business volume, Need to do the main part of dynamic expansion.
NFC Testing Solutions
For each component in Clearwater, define a kubernetes service, each of which contains one or more pods that can be scaled horizontally by increasing the number of pods.
The list of services in Kubernetes is as follows:
The pod list in Kubernetes is shown in. During the test, the number of concurrent users is increased by the Clearwater component based on Sipp, the system pressure increases, and the pod count in the Sprout service is increased.
In order to realize the dynamic expansion of pod, on the one hand, using the replication controllers of kubernetes itself to increase or decrease the number of pod copies according to load, on the other hand, Clearwater Sprout is a micro service architecture, distributed, stateless, Separation of data and applications allows the cluster to share the database with each container in the cluster.
When the number of concurrent users is 300, the number of Sprout module pod is generally two or three, when the number of concurrent users increased to 2000, the number of Sprout module pod reached 11, if continue to increase the number of concurrent users to 3,000, Sprout module pod number reached about 15.
Ct-it Fusion Test Solution
In big data applications, the lightweight, high-performance, fast-start features of the container also have advantages, while the CT and IT business integration deployment is also a trend in the future, in this test, we are on the same container platform, the deployment of IMS network element and big data applications.
In the low load ratio of CT application, the system is mainly used for the calculation of big data, and the proportion of CPU resources is higher in big Data application. As shown in the resource monitoring interface.
When the CT business volume increases, the system load increases, in order to meet the online CT business requirements, increase the number of CT business containers, while reducing the use of large data offline business resources. The resource usage changes as shown.
Summarize
As a lightweight virtualization technology, containers can reduce system overhead and improve system performance.
- ct/it converged deployment: CT and IT applications are deployed in the same platform, sharing resources, and improving resource utilization.
- Unified Resource Scheduling: Break application silos, unify construction, unify operation and maintenance, reduce TCO
- Improved system performance: Use of container technology, lightweight virtualization, lower overhead
- MicroServices Architecture: Distributed, stateless cloud-native applications that enable separation of data
- Elastic scaling, Automatic expansion: Support a custom resource scheduling strategy to achieve dynamic scheduling of resources based on traffic changes.
Q&a
Q: Is there a bottleneck in network forwarding? Have you tried some acceleration techniques, like DPDK?
A: Because it is the control surface network element, there is no bottleneck, no use of DPDK, we have tested the virtual machine before, because the characteristics of the network element, is not configured DPDK, the current test on the performance of no problem.
Q: Is the database in the container now?
A: Yes, but it's also a database cluster, and we're doing the work of separating the databases.
Q: How do your container networks implement NFV interconnection, as your measured containers run in a virtual machine environment?
A: The container is not based on the virtual machine environment, because it is the control surface network element, the performance has not encountered the bottleneck.
Q: Container does not fit all, and virtual machine, physical machine coexistence may be a period of time a normal, we have not analyzed which business is suitable for containerized, which business should not be changed? And how to realize the interoperability between the container and the virtual machine application when using Kubernetes as the container's orchestration scheduling tool?
A: In fact, business should not change not only the technical problems, but also business vendors are willing to change the relevant, after all, some legacy business has been in the current network operation for a long time, it is difficult to container in the short term. In addition, we do encounter a business that requires kernel optimization. We have not yet done the interoperability of containers and virtual machines in our system.
Q: Traditional CT network is not only IMS, for CS, PS,SDN/NFV, do you have any ideas?
A:ims let us containerized the first attempt, the other network elements are also being studied, follow-up will do the relevant tests, if interested, also welcome to participate in the test behind us.
Q: Do you have a resource pool with functions such as routers, security groups, and load balancing? Routers, can a load balancer be assigned a public network IP?
A: Refers to the resource pool of the container or the private cloud resource pool we already have, the existing private cloud resource pool is there, but the container is not the prototype system yet.
Q: What else is an open source project like Clearwater?
A:NFV aspect of, if is ims,clearwater industry uses more, NFV other net element open source project many.
Q: What virtualization Technology Support does your virtual machine use at the bottom?
A: This demo, the bottom of the use of Docker, our private cloud resource pool has VMware, KVM and Xen.
The above content is organized according to the August 16, 2016 night group sharing content. Share people
Ma Yihui, a private cloud project manager at the Research Institute of China Mobile Communications, previously worked as a research engineer at VMware Research and Development Center. has been concerned about and committed to virtualization, containers, cloud computing and other related fields. Dockone Weekly will organize the technology to share, welcome interested students add: Liyingjiesz, into group participation, you want to listen to the topic or want to share the topic can give us a message.
