Importance of a Flat Network Architecture in Cloud Implementation
Executive Summary
As adoption of server virtualization gains momentum, many customers are bracing themselves for the end goal – building a self-service private cloud. However, changes to the network infrastructure have not kept pace with the level of transformation happening in server virtualization. As a result, the network often becomes the performance bottleneck, preventing customers from fully realizing the benefits of their virtualized infrastructure and private cloud.
Data center networks need to evolve to be more virtualized, flat, intelligent and automated. The network fabric must mirror the transformation in the computer architecture for organizations to fully realize the benefits of business agility, operational efficiencies and lower costs.
In August and September 2011, Brocade commissioned Frost & Sullivan to evaluate the network implications of virtualization and cloud computing. Frost & Sullivan surveyed 328 IT decision-makers from enterprises across Asia Pacific. This research reveals that organizations need to address network challenges in a more pro-active manner rather than as an afterthought.
Some of the key findings of the study include :
- Server virtualization has crossed the inflection point with 67% of organizations adopting it. 46% of organizations that have virtualized are running production environments on virtual machines (VM).
- 35% of organizations are using cloud in one form or the other. Private cloud is the most popular deployment model with 40% of cloud users using this delivery model.
- Network virtualization, flattening the data center network architecture and simplifying network management are top priorities for IT decision makers.
- Ethernet fabric architecture is becoming critical due to its scalability, flatness and efficiency.
To better support their virtualization and cloud infrastructure, customers need to:
- Virtualize the network layer for seamless creation of resource pools.
- Optimize the network infrastructure through flat architecture, protocol enhancement and open frameworks.
- Automate the network to be VM aware and enable elastic scaling with automated provisioning of resources.
- Pro-actively monitor and manage the network to increase availability and reduce operational costs.
State of Cloud Computing in As ia Pacific
The IT industry is undergoing a paradigm shift in the way computing resources are being procured and delivered. Cloud computing is emerging as a key area of focus for IT decision makers due to its ability to save costs, increase business agility and deliver IT in an on-demand model. In the recent past, interest in cloud computing has grown exponentially in Asia Pacific and is increasingly a focus area for CXOs. While SaaS adoption has been steadily growing in the past decade, the adoption of IaaS and PaaS has seen a huge spurt in the past 12 months.
Among the technology factors that will serve as enablers for the delivery of cloud services, virtualization has the greatest impact since it provides a foundation for delivering cloud solutions. Virtualization helps to build clouds by creating logical instances that can be provisioned on-demand.
Customers have a choice of different deployment models in the cloud, namely, public, private, and hybrid clouds. It is important to understand the difference between each one of these models.
The adoption of private clouds has grown significantly in the last few months. This is due to the ability of private clouds to offer several benefits of public clouds while addressing the disadvantages associated with this kind of deployment.
Out of 35% of organizations who are using cloud, 14% are using private cloud, 9% are using hybrid cloud and 12% are using public cloud.
However, a private cloud rollout is often a complex process that can involve significant investment in technology, processes and people.
This white paper examines some of the important steps in implementing a private cloud with particular emphasis on the network implications of a cloud rollout.
Cloud Implementation Phases
Figure 1: (Cloud Implementation Phases)
1.) Consolidate
Organizations can generate substantial efficiencies by consolidating their infrastructure, primarily through various virtualization techniques. Virtualization is the first step towards cloud adoption as it creates seamless logical pools of resources and increases asset utilization. It makes applications independent of the underlying hardware infrastructure, thereby allowing flexible service creation. When implemented properly, virtualization can lead to more efficient provisioning of resources, increased flexibility and lower operational costs. While server virtualization is the most known type of virtualization, it is equally important to consider the possibility of virtualizing storage systems, applications, network infrastructure and end points, such as desktop computers – each of which can help with consolidation efforts.
A recent Frost & Sullivan survey1 of 328 enterprise IT decision makers in Asia Pacific reveals that over two thirds of organizations have virtualized their server infrastructure.
According to Frost & Sullivan, 67% of organizations in the APAC have performed some forms of virtualization.
Although early adopters used server virtualization primarily for test and development environments, more customers today are moving production workloads to a virtual machine. In a clear sign of growing maturity of server virtualization, slightly less than half (46%) of customers who run virtual workloads, use server virtualization on their production systems.
This transition is significant since it shows the growing maturity of the virtualization market, and is an important first step in the move towards building a private cloud. However, while server virtualization can solve many challenges of traditional server computing, it poses new issues around the network and storage infrastructure. Existing data center network architectures that are reliant on spanning tree protocol (STP) with hierarchical three-tier architecture may not be well suited for virtual workloads and growing east-west traffic. Therefore, creating a data center architecture that is simple, secure, flat and virtualized is critical. By virtualizing the network fabric, customers can support their server virtualization and cloud initiatives much more effectively. It helps them achieve the benefits of higher operational efficiencies and lower costs.
This transition is significant since it shows the growing maturity of the virtualization market, and is an important first step in the move towards building a private cloud. However, while server virtualization can solve many challenges of traditional server computing, it poses new issues around the network and storage infrastructure. Existing data center network architectures that are reliant on spanning tree protocol (STP) with hierarchical three-tier architecture may not be well suited for virtual workloads and growing east-west traffic. Therefore, creating a data center architecture that is simple, secure, flat and virtualized is critical. By virtualizing the network fabric, customers can support their server virtualization and cloud initiatives much more effectively. It helps them achieve the benefits of higher operational efficiencies and lower costs.
Network virtualization is critical in order to address many of the challenges posed by server virtualization. Network virtualization is not an entirely new concept since the usage of VPNs (Virtual private network) and vLAN s (Virtual local area network) has been in existence for a long time. Network virtualization enables network aggregation and provisioning, combining different physical networks into a single virtual network, or breaking a physical network into multiple virtual networks that are isolated from each other. A virtual network treats all hardware and software in the network as a unified collection of resources, which can be accessed regardless of physical boundaries.
In physical environments, adding switch ports requires additional cabling, connections and configuration. In a virtual environment, logical switch ports are created and abstracted from the underlying physical ports. It therefore streamlines network operations, simplifies management of network resources, and lowers costs through efficient resource usage.
2.) Optimize
Consolidation of IT infrastructure delivers efficiencies through increased asset utilization. However, consolidation needs to be followed by optimizing the underlying storage, network and service management layers so that they are aligned to supporting the needs of a virtual infrastructure.
Frost & Sullivan research indicates that there are three key changes required in the network architecture that will help data centers to cope better with virtualization and cloud computing.
Figure 2: (Improving Data center network efficiency)
Â
Protocol Enhancement: Firstly, organizations need to look at replacing STP with new protocols such as transparent interconnection with lots of links (TRILL). Once it is ready to be widely adopted, TRILL can free up more layer-2 paths, which in turn can increase bandwidth availability.
Architecture Redesign: Most of the existing data centers are based on the three-tier architecture, involving access, aggregation, and core layers. This architecture proves to be inefficient in a virtualized environment, as it is not optimized for server-to-server and server-to-storage traffic. Traffic continues to flow through different tiers, which adds to latency with each hop, thereby affecting the performance of real-time applications. In recent years, there has been a significant improvement in the performance of switches, which enables a shift to two-tier architecture by eliminating the aggregation layer. Furthermore, the elimination of a tier reduces the number of switches and cables required in a data center network, leading to lower costs, both operational and capital.
Adoption of Open Frameworks: With the rising heterogeneity in data center networks, it is critical that customers choose solutions that adhere to open industry standards. This future proofs their investments. In order to fully optimize the network for it to be cloud ready, customers deem the following factors to be most important:
Figure 3: (Changes needed in the data center network for it to be cloud ready)
Â
Network Optimization, through usage of Ethernet fabric architecture, is an important step in ensuring that the network operates at peak efficiency. It creates a resilient and flexible network infrastructure that can seamlessly scale to support dynamic business needs and allows administrators to maintain agreed-upon service level agreements (SLA ).
3.) Automate
The true promise of cloud – business agility and on-demand provisioning of resources – can only be delivered if workflows are automated. Automation can convert many of the laborious and time-consuming manual processes into seamless workflows, thereby reducing repetitive processes.
The following chart outlines the various phases in Cloud automation.
Figure 4: (Various phases in cloud automation)
Network automation is an important component of overall cloud automation. It needs to support on-demand provisioning of resources and workflow automation.
4.) Manage
Maintaining high availability and service levels of private clouds is an important imperative for both service providers and in-house IT teams. Private clouds are complex to monitor due to the virtualized nature of core infrastructure resources (compute, storage, and networking) which are always in a state of flux.
However, it is important to reduce potential performance bottlenecks before they impact business operations through pro-active monitoring and alert management. A pro-active early warning system can accelerate troubleshooting and allow organizations to further improve and fine-tune the monitored systems. While monitoring is an important facet of management, the ability to isolate, analyze and report traffic patterns to streamline network operations will be an important evolutionary step for customers to consider.
Figure 5: (Various phases in cloud management)
Brocade’s value proposition in enabling a cloud ready network
In 2010, Brocade introduced its vision, code named, Brocade One, a cloud and data centre architecture, with Virtual Cluster Switching (VCS) as the core technology for building large, high-performance and flat layer 2 data centre fabrics to better support server virtualization and cloud computing.
Recommendation
Data center network architecture needs to evolve and keep pace with the changes in server virtualization. The transition to private cloud architectures will only succeed if organizations streamline and orchestrate technologies, processes and people. The network is an important element in the cloud architecture and Frost & Sullivan recommends customers to focus on the following factors and consider them while building their clouds.
