Cloud computing cannot afford the next Generation data Center network drawing thoroughfare

The new Generation data center is a high density, large capacity of network applications, the network bandwidth, port density, security protection and management have put forward a severe test.

We see that the cloud computing data center than the traditional data center on the requirements of the network has changed a lot, which mainly includes server-server traffic into the mainstream, but also requires two-tier traffic; The increase in the number of physical servers and virtual machines within the site results in a two-tier topology; expansion, Disaster preparedness and VM migration require that the data center multi-site between the two tiers of interoperability, data center multi-site selection problem by the sophomore layer of interaction more complex and so on.

Virtual Machine Interconnection

From the current technology, the ithttp://www.aliyun.com/zixun/aggregation/13748.html "> Infrastructure of the two basic servers and storage in terms of performance and virtualization features such as relatively perfect, Comparatively speaking, the challenge of the network partial Pro is bigger, it is possible to become the bottleneck of cloud computing application. H3C that creating a cloud-ready network needs to address several issues:

• Cluster computing, the need for port support 10GE high-density access, 40/100g, single machine with 10000 server access capabilities, multiple single domain virtualization, to achieve cross machine integration, to be able to cross domain virtualization, to achieve cross-data center integration, and support security virtualization

• Virtual computing, realizing virtual machine perception through eVB, and achieving virtualization security,

• Lossless switching, support for CEE Fusion enhanced Ethernet, and FCOE to achieve data center IO fusion

• Automated management, automated configuration and automated scheduling

"Deploying High-capacity network platforms, solving virtual machine access problems, and building a sophomore plane network to adapt to resource adjustments is an urgent need for the network to meet the infrastructure effectiveness of cloud computing," Hongliang, Minister of Development, H3C data center, said in an interview with ZDNet reporters. ”

From the number of servers, the data center presents an obvious development law: The data center of the mass server scale will be the mainstream of development. At the same time, with the increase of the number of computing cores, the bandwidth consumed by the single server is increasing synchronously, and the hardware acceleration technology of 10GE Network card is approaching the speed of the throughput of the server.

On the market, we see that in order to solve the high density access problem brought by server virtualization, H3C has introduced a new generation of S5820 high-density gigabit switches that provide 48个万 gigabit SFP ports and 4 40GE ports, and each 40GE port can be easily extended to 4 Wan ports. H3C launched the S12500 switch to support the future 40GE and 100GE Ethernet standards, facing the next generation of data center burst traffic, the use of "distributed access caching" technology, can achieve data 200ms cache, to meet the data center, High-performance Computing and other network burst traffic requirements.

With the final finalization of the 40g/100g standard, it can be foreseen that a 40g/100g interface storm will soon come.

After server virtualization, how to connect the network, how to realize the management and control of the virtual machine in the network layer, will be the new problem that the network faces under the cloud computing environment. In response to this issue, HP and H3C presented the VEPA programme to the IEEE eVB organization in 2009. Hongliang pointed out that the Huasan Vepa make the boundary gradually clear, the vswitch understood as a network cable, one into a channel into a direct access to the hardware switch, so that the vswitch complexity as much as possible. The main switching task is to do it on the hardware platform, which can make the network clear again based on Vepa.

Sophomore Layer Planar Network interconnection

With the development of the application of the upper layer, the virtualization technology and the large-scale centralization technology are widely used in the enterprise it, as the network of the underlying infrastructure, it also enters a new round of technological innovation period.

In a cloud computing environment, computing and storage resources are typically built in the form of resource pools. In-pool virtual machine migration and cluster are two typical application models, both of which require two-tier network support. With the continuous expansion of cloud computing resource pool, the scope of the two-tier network is gradually expanding.

Hongliang points out that "the migration of virtual machines requires the construction of a planar network, such as Huasan's Irf2,trill, Cisco's Fabricpath, to be prepared for this." ”

With the goal of simplifying the network logic, integrating the physical nodes and applying the fast change of the upper layer, the IRF2 realizes the simplicity of the IT network operation, and changes the cumbersome rules of the traditional network planning and design.

IRF2 virtualization Technology Not only simplifies multiple physical devices into a single logical device, moreover, the multiple link connections between the layers of the network become the direct connection between the two logical devices, so multiple physical links can be aggregated across the device to become a logical link, This avoids loop problems caused by multiple links.

Hongliang told reporters that for the two-layer loop problem, the H3C uses IRF2 technology to virtual multiple network devices into one, through the link bundled across the device to eliminate the network loop, while these devices as a unified management, configuration and use.

(Responsible editor: The good of the Legacy)