CloudFabric for Cloud Evolution
The extensive use of cloud computing, Big Data, and mobile Internet is accelerating the launch of new services and creating more data center traffic. Huawei CloudFabric is an application-driven cloud data center network solution capable of quickly responding to fast-changing services. It enables enterprises to build simple, open, and scalable cloud data center networks that will support long-term cloud evolution. CloudFabric provides multiple capabilities, including scalable interconnection, automated deployment, FabricInsight refined O&M, and an open ecosystem. These features enable real-time service provisioning, rapid fault location, active fault recovery, and quick integration of application systems.
Huawei’s cloud data center network solution simplifies the construction of data center networks by combining industry-leading data center switches with comprehensive transmission, routing, security, and network management products. Switches include the flagship CloudEngine 12800 series core switches that provide the industry’s highest performance: CloudEngine 8800/7800/6800/5800 series fixed switches, and the CloudEngine 1800V virtual switch. Huawei switches, together with the Huawei Agile Controller, are the foundation of modern, SDN-based, data center networks.
Separating virtual and physical resources
Cloud services require computing, storage, and network resources, but the separation of virtual (IT) data center resources and physical (network) resources makes it difficult to unify resource allocation, fault association, and diagnosis, and hinders the deployment of cloud services. Without advanced network resources, the efficiency of the physical network will continue to lag behind the virtual IT network.
For example, consider the problems of an aeronautical enterprise. Due to frequent service changes, the IT system management department dispatches 1,000 tasks to the network management department every month. However, the network management department can complete only 70 percent of the tasks in a month. Gradually, as more and more tasks are left unfinished, service development is slowed. Correspondingly, fault detection is slowed. The IT system can detect some faults but cannot accurately locate failure points in a timely manner because it cannot associate them with information from the physical network. These problems urgently need solutions:
The physical network needs to be quickly and automatically deployed in the same way as IT resources
The physical network must be a part of the cloud, and physical resources must be allocated uniformly with IT resources, including computing and storage data
Physical network and virtual network resources (such as vSwitch from VMware) should share a unified view to implement comprehensive, unified management
The physical network should evolve stably and smoothly to meet the requirements listed above, and support cloud service development
With CloudFabric, enterprises can build a fully meshed cloud service system to connect cloud applications, cloud platforms, and various ICT resources on data center networks that function as information exchange centers.
Figure 1. CloudFabric architecture
CloudFabric architecture, a service-driven solution
The CloudFabric Solution uses a service-oriented, open architecture to uniformly allocate ICT resources through the Agile Controller and cloud platforms. The Agile Controller can connect multiple mainstream cloud platforms to provide a cloud application-oriented network. Users can uniformly schedule network, computing, and storage resources on demand.
Because it is service-centric, the CloudFabric Solution permits IT administrators to easily schedule network resources and implement cloud-based network migration. IT administrators can use the service language to define and adjust network requirements. The Agile Controller, meanwhile, enables network resources to be migrated on demand.
Figure 2. Application framework and logical and physical networks deployed as IT resources
Data center ecosystem
Huawei is actively developing a cloud computing data center ecosystem that adheres to its customer-centric philosophy. The CloudFabric Solution has been used by globally known IT vendors and standards organizations to connect to mainstream cloud platforms such as those developed by VMware and Microsoft. The ecosystem includes OpenStack and Huawei’s FusionSphere, which centrally manage network and IT resources.
Figure 3. The CloudFabric ecosystem includes OpenStack, VMware, Microsoft, and Huawei FusionSphere
Components of Interconnected Cloud Platforms
Huawei’s Agile Controller is an element of interconnected cloud platforms: it integrates functions, including physical and virtual network management and automatic service-chain orchestration. The controller also works with cloud platforms to dynamically schedule network resources, making it easier and quicker to deploy cloud services. Its comprehensive openness and programmability consists of:
The Agile Controller: Provides northbound interfaces to connect to mainstream cloud platforms and virtualization platforms for delivery and adjustment of network policies. It also provides southbound interfaces to manage physical and virtual networks (including third-party devices) based on open API technology and OpenFlow.
Device-level openness: Permits customers to flexibly select devices. CloudEngine series data center switches support a large variety of open APIs plus standard OpenFlow, and can directly connect to mainstream cloud platforms.
Multiple cloud connect solutions: Huawei has worked with mainstream IT vendors to provide solutions that include Virtual Machine (VM) network policy migration, VXLAN/NVGRE virtualized gateways, and ICT collaborative management.
By connecting ICT resources, CloudFabric provides IT and network administrators a global view of data center networks that displays physical and virtual network resources. Administrators are able to assess global network resource distribution, resource usage, and fault status in real time. When network faults occur, the Agile Controller can quickly and accurately locate failure points and adjust the physical and virtual networks accordingly. It can also associate applications with the logical network. In this way, upper-layer applications are aware of network faults, ensuring quick service optimization.
Figure 4. Separated vs. mutually-aware physical and virtual networks
The CloudFabric Solution decouples logical networks from physical networks so that IT administrators can focus on logical networks and flexibly customize logical network policies based on service requirements. Logical networks shield the differences between various physical and virtual networks and networking technologies. As a result, CloudFabric is compatible with any physical network model, and customers avoid reinvestment caused by replacing devices and reduce TCO.
Figure 5. The Agile Controller and VMware vCenter enable on-demand policy migration
Huawei and VMware collaborate on cloud data centers
The Agile Controller, which controls the data center network, connects to the VMware vCenter through an open API. When vCenter starts the process of migrating virtual machines, the Agile Controller migrates the associated network policy along with VMs and delivers the configuration to CloudEngine series switches.
The Huawei Agile Controller and VMware vCenter also work together to schedule ICT resources for migrating VMs, enabling on-demand policy migration. This in-service policy migration arrangement, jointly developed by Huawei and VMware, has the following characteristics:
The Agile Controller delivers network policies using high-speed RADIUS interfaces, which increase the speed of policy deployment to 10 to 20 times the industry average. This high-speed policy deployment can keep pace with the migration of a large number of virtual machines.
Huaweiʼs large Layer 2 network solutions (such as TRILL and EVN) enable VMware virtual machines to migrate freely within a data center or between data centers, making cloud service migration more flexible.
In addition to the VM migration solution, Huawei and VMware cooperate in many other fields. Huawei Agile Controller is extensively integrated with the VMware vCloud and NSX network virtualization platforms to provide a more intelligent, efficient cloud network solution.
Figure 6. Huawei and Microsoft cooperate in solution marketing and standardization for data centers
Huawei and Microsoft in the data center
Huawei and Microsoft, both leading ICT vendors, formed a strategic partnership to develop new technologies for cloud connect and overlay solutions. The two ICT giants will also cooperate in Huawei CloudEngine series data center switches and Microsoft Cloud OS (Windows Server 2012 R2 and System Center 2012 R2) are closely coupled to create uniform management of ICT resources. CloudEngine series data center switches and Microsoft Cloud OS (Windows Server 2012 R2 and System Center 2012 R2) are closely coupled to create uniform management of ICT resources.
Figure 7. Huawei’s CloudEngine series data center switches provide a variety of open API interfaces
The CloudEngine series data center switches are built on Huawei's next-generation network operating system VRP8 and integrate the Open Programmable System (OPS) module that provides a variety of open API interfaces. All CloudEngine series switches (CE12800/6800/5800) have passed Open Management Infrastructure (OMI) certification for the Microsoft Cloud OS and are integrated into the operating system.
Figure 8. Huawei’s CloudEngine series data center switches provide a variety of open API interfaces
The Microsoft Hyper-V network virtualization platform and Huawei CloudEngine 12800 series data center switches work together to offer a high-performance, flexible Hybrid Fabric solution.
The Microsoft vSwitch — with the Network Virtualization using Generic Routing Encapsulation (NVGRE) protocol — can build an overlay-virtualized network. Huawei CloudEngine 12800 series core switches can be used as hardware NVGRE gateway devices on the overlay network.
The NVGRE gateways implement high-speed communication between Microsoft virtual machines and campus/branch users, or between Microsoft virtual machines and other vendors' virtual machines and physical servers. Huawei and Microsoft can jointly build a Hybrid Fabric network that combines the high performance of hardware and the flexibility of software. This network is highly scalable and can support up to 16 million tenants.
Microsoft System Center can manage Huawei CloudEngine series switches and Microsoft vSwitches.
Huawei/Microsoft Success Story
Huawei and Microsoft cooperate to explore the data center market and have provided cloud data center solutions for multiple customers, for example, China Mobile (Guangdong).
The data center project of China Mobile (Guangdong) is a successful implementation of the Hybrid Fabric solution developed by Huawei and Microsoft. This solution includes two data centers in Guangzhou and Shenzhen. Huawei CE12800 switches are used as NVGRE hardware gateways and enable a large number of tenants to access Microsoft virtual machines in the NVGRE domain. The switches also implement in-service migration of Microsoft virtual machines between the two data centers.
Figure 9 China Mobile (Guangdong) data centers in Guangzhou and Shenzhen
CloudFabric and SDN Build Next-Generation Networks
Today, 10 GE servers dominate data centers because of exponential traffic growth created by increasing Big Data applications. By the end of 2020, 40 GE/100 GE servers will overtake them. Internal data center traffic is also surging because interfaces connecting data center networks have evolved to 40 GE/100 GE. Because of these trends, data centers require Huawei’s service-centric, open architecture that provides non-blocking, high-speed packet forwarding.
To summarize: The Huawei CloudFabric Data Center Network Solution uses SDN technology to build next-generation cloud data center networks. Core components include industry-leading CloudEngine series data center switches, Agile Controller, and a variety of transmission, routing, security, and network management products. These products together create a one-stop data center network solution.