--H3C solution for cloud interconnection across data center two-tier network

The following three types of interconnected links are typically deployed between data centers, and each interconnect link carries different data and implements different functions, as shown in the following illustration:

Three-layer interconnection. Also known as Data Center front-end network interconnection, the so-called "front-end network" refers to the data center for the Enterprise Park network or Enterprise WAN export. The front-end networks of different data centers (main center, Disaster preparedness Center) are interconnected through IP technology, and the campus or branch clients access each data center through the front-end network. When a disaster occurs in the primary datacenter, the front-end network is fast converging and the client accesses the disaster Preparedness Center to ensure business continuity.

Two-layer interconnection. Also known as Data Center Server network interconnection. In different data center Server network access layer, construct a sophomore network (VLAN) across the data center, to meet the requirements of two-tier network access in scenarios such as server cluster or virtual machine dynamic migration.

San Interconnect. Also known as back-end storage network interconnection. With the aid of transmission technology (DWDM, SDH, etc.), the data replication of disk array between main center and disaster standby Center is realized.

In traditional data center Server area Network design, three-layer interconnection and San interconnection are generally available, but usually do not have two-layer interconnection channel. In the Cloud Computing data Center interconnection, in order to achieve the data center resource scheduling, business cluster and virtual machine migration, the network layer to ensure that the different data center VLAN two-tier interoperability. However, to achieve two tiers of interoperability across the data center network, traditional technology solutions face the following challenges:

1.1 City Double Center

In the same city double center construction mode, the double center usually uses the bare fiber or the DWDM link interconnection, therefore the link quality and the link bandwidth is high, may satisfy the business double live and the load sharing demand very well. But in this environment, if the VLAN of two data centers is opened directly, it will bring the docking problem of spanning tree, and the spanning tree Protocol of the whole network is more complicated.

In this environment, using H3C IRF2 technology, the End-to-end deployment of IRF virtualization in two data centers eliminates network two-tier loops and spanning tree protocols within two data centers. When two data centers are docked by cross equipment link bundle technology, not only will not introduce the new interconnection loop, but also can improve the interconnect bandwidth and network reliability. As shown in the following illustration:

1.2 Three centres

The three central construction mode of the two places, the remote data center interconnection usually does not have bare fiber or DWDM link, generally using leased operators wide-area transmission network (such as SDH, etc.), with operator wide-area transmission network can not directly achieve two-tier VLAN interoperability. In this case, it is generally necessary to use tunneling technology to achieve two-tier interoperability, according to the wide-area core transmission network architecture, across the wide area of the cloud data center two-tier interoperability is divided into MPLS core network and IP core network two schemes.

1. MPLS Core Network

In the MPLS core network environment, we can directly deploy VLL and VPLS standard technology, build two-layer VPN tunnel based on MPLS Core network, realize point-to-point and point-to-point two-layer network interconnection.

2. IP Core Network

In the IP core network environment, cannot directly deploy VPLS,H3C original EVI new technology, can build the tunnel through the GRE in the IP core NET, realizes the point-to-point and the point-to-point two layer network interconnection. This technology has no special requirements for IP core network transmission devices.

1.3 One-level multi-center

For the two data centers in the city of the point-to-point bare fiber (or DWDM) interconnection, because the Internet is relatively simple, will not introduce a new interconnected circuit. However, if the need for the interconnection of the same city cloud data center reached three and above, the use of raw fiber common mode of interconnection will inevitably cause the Internet loop, in this environment, H3C recommended Hub-spoke star model, as shown in the following figure:

In the Hub-spoke network model, the hub node is the key of the whole network, the hub node failure will cause all cloud center Internet to be interrupted. H3C Remote IRF technology can be used in actual deployment, the two devices of hub node are placed in different data centers respectively, and the IRF virtualization of Cross Data center is realized through bare fiber or DWDM, and the hub node is guaranteed to be high and reliable on the basis of realizing the unified management of hub node.