To ensure the success of cloud computing, enterprises and cloud operators must ensure their server, storage, network operations, and other critical building blocks are cloud-ready.
To harness the considerable advantages of cloud computing, networks must be equipped with five essential elements high-bandwidth/low-latency switching, convergence to Ethernet, agile networking for virtual workloads, scalable management and advanced energy efficiency. New approaches to networking must move beyond commodity switches to extend the performance and capabilities of cloud-ready networks from blade server to switch to SAN that will enable the scale-out performance and economies promised by cloud computing.
High Bandwidth with Low Latency
In the past, adding bandwidth meant adding more Gigabit Ethernet connections. Today, the rapidly rising server utilisation and network traffic levels fueled by technology advances such as multi-core processors, virtualisation, and the convergence of storage and data networks have made that strategy impractical. Scaling the number of Gigabit Ethernet connections to the levels necessary for cloud computing would be too costly and require too many ports to manage efficiently.
With 10 Gigabit Ethernet (GbE) now a viable technology and low-latency 10GbE switches available for around $500 per port, cloud computing becomes a realistic alternative. Cloud computing solutions can be practical for users and profitable for service providers if they offer 10GbE performance and low latency to support high application performance and server utilization.
Converged Communications and Storage
By nature, cloud computing networks will carry a massive amount of data. Therefore, convergence will be vital to cost-effectively building out, maintaining, and managing a cloud computing infrastructure. No one wants separate LAN and SAN infrastructures when one will do the job. And with the availability of Network Attached Storage (NAS), iSCSI and Fibre Channel over Ethernet (FCoE), that network is certain to be Ethernet. Also vital: FCoE must be deployed with loss-less characteristics to ensure that storage traffic is delivered reliably and in a timely manner. A loss-less implementation will also add the benefit of increasing TCP/IP traffic efficiency.
Mobile Virtual Machines
One of the main advantages of cloud computing is on-demand access to resources, and virtualisation plays a key role in providing those resources. This advantage of on-demand resources can be greatly magnified with the ability to move virtual machines between physical servers while applications continue to run. An infrastructure with mobile, active virtual machines can respond to new requirements much more quickly and cost-effectively. Cloud computing users can gain even greater advantages from mobile virtual machines when they can be moved not only within a cloud, but over greater distances to connect multiple clouds. Movement between clouds enables applications such as disaster recovery and data replication.
Moving running virtual machines has traditionally been a complex task because of the time-consuming requirements of moving the network-level policies associated with each virtual machine to its new location. These policies govern factors such as security and Quality of Service (QoS). The policies differ based on the users and applications, making their movement with the virtual machine critical. Without a virtualisation-aware network, moving policies along with virtual machines requires extensive manual configuration and carries a significant opportunity for misconfiguration and security breaches if not implemented correctly.
In the fluid cloud environment, administrators need tools that can manage mobile virtual machines with their associated policies, and scale to potentially tens of thousands of physical servers. Such tools and the ability to automate mobile virtual machine management will be critical to administrator productivity.
Scalable Management Tools
Outside the realm of virtual machine management, administrators also need basic management tools that can scale to cloud computing proportions. Server-by-server management is simply not practical when a network is handling a trillion packets per second. All management tools must be scalable and replicable to minimise administrator effort. Real-time metrics, ready navigation through logical resource groupings, and the ability to quickly drill down to see problem areas are important features of managing cloud computing networks.
Advanced Energy Efficiency
Power is one of the biggest expenses facing data centre managers today. New environmental mandates combined with rising energy costs and demand are forcing administrators to focus on green initiatives. As clouds grow to thousands of servers and beyond, per-component power savings will be magnified into vital budget reductions.