Data Center Directions
, Regular Visitor
Data Center Directions
A Network Fabric for New Applications in the Data Center
Apr 24, 2014

The Network Challenge with Deploying Applications
Thanks to virtualization organizations can bring up new applications and services quickly. Unfortunately, many data center networks don’t let you fully capitalize on the business agility that virtualization and modern application architectures provide. Traditional network architectures are too slow and too cumbersome to configure. For true agility, enterprises need a high-performance, low-latency fabric network that can be managed like a single, logical switch.


Today’s mid-sized and large-scale data centers are built with high-performance blade and rack servers, which typically run multiple VMs, which in turn run increasingly modular, web- and cloud-based applications. These modern applications are driving increased traffic levels and different traffic patterns, which place specific demands on the data center network.


Critical business applications such as enterprise resource planning (ERP) and customer relationship management (CRM) are divided into multiple modules. As a result, relatively simple operations such as entering an order can trigger 100 or more individual transactions, each imposing its own latency. Some applications, including eCommerce applications, even behave dynamically, spinning up new instances or migrating workloads in response to traffic loads.


Their distributed nature means modern applications are spread across racks of servers, each served by multiple switches. The applications generate a tremendous amount of server-to-server, or east-west, traffic as the various modules communicate with one another. Multi-tier network architectures aren’t well matched to modern applications. They force this east-west traffic to first travel north and south, up and down the network tree, before arriving at its ultimate destination, adding significant latency that can cause application performance to degrade under load.


Latency adds up quickly in a multi-tier architecture. A few milliseconds at each hop become seconds of latency per transaction, which seriously impacts application performance—and user experience. Research has shown that even two seconds of delay prompts a majority of customers on eCommerce web sites to abandon their shopping carts .


Multi-tier network architectures are also too complex and expensive to operate. Each time a new application is deployed, network administrators must configure the network device by device, driving up operations overhead and slowing down application roll outs. With multi-tier network architectures, administrators have too many switches to manage, too many management tools to learn, and too many manual processes to execute.


To support today’s application environment, enterprises need a high-performance, low-latency data center network that costs less to operate, automates management functions and simplifies change. Over the past few years, fabric technologies have become popular solutions for implementing efficient data center networks, reducing traditional three-tier networks to flat systems featuring predictable performance and deterministic low latency along with plug-and-play deployment and the ease of managing a single logical switch.


Juniper's Virtual Chassis Fabric is the Answer

Juniper’s Virtual Chassis Fabric technology delivers a high-speed, flat data center networking solution with the deterministic low latency and predictable performance that applications need, and the low operational overhead of a single logical device.


Virtual Chassis Fabric technology delivers the following features and benefits:


A single entity to manage: Virtual Chassis Fabric technology gives IT a single logical point of management for all interconnected switches, presenting one management IP address that is accessible from any member of the fabric. One spine switch acts as the Master device, handling all communications and control for the fabric.


Zero-touch, plug-and-play provisioning: Juniper designed Virtual Chassis Fabric configurations to be much easier to deploy. By factory default, supported devices automatically join the fabric and the Master RE can push out configuration information to other switches, automatically provisioning spine and leaf nodes.


Smart network management: Integration with the QFX5100, Insight Technology for Analytics helps IT make better network design decisions and identify network hotspots by providing dynamic buffer utilization monitoring and reporting.


L2 and L3 capabilities on the same device: Virtual Chassis Fabric provides rich L2 and L3 functionality, which gives IT the flexibility to use the connectivity method that works best for the applications being supported.


Resiliency and high availability: Virtual Chassis Fabric technology has robust redundancy features that ensure reliability and uptime as well as non-disruptive maintenance operations including redundant routing engines, data plane redundancy with active-active uplink forwarding, and uplink redundancy as well as server multi-homing.


Topology-Independent In-Service Software Upgrade (TISSU): An industry first, TISSU allows for in-service software upgrades on individual top-of-rack switches—no redundant switch topology is necessary.


Investment protection: Juniper designed its new fabric architecture with flexibility and investment protection in mind. For example, customers who have deployed the EX4300 or the QFX3500, QFX3600 or QFX5100 switches in a Virtual Chassis configuration can easily migrate to Virtual Chassis Fabric.


Management plane: The management plane enables up to 20 switches to be managed as a single device, greatly simplifying all aspects of management, from initial configuration and deployment to ongoing operations and upgrades.


Control plane: The Virtual Chassis Fabric technology features a fully redundant, integrated control plane and up to four integrated, redundant routing engines that provide centralized control for network ports.


Data Plane: The Virtual Chassis Fabric data plane provides low latency and predictable performance for both L2 and L3 by supporting local switching on all ports and having all fabric links operating in active-active mode, with traffic load balanced on all links.



Virtual Chassis Fabric is Right for Your Organization
Virtual Chassis Fabric was designed to address the requirements of financial services, federal government, healthcare and other organizations that are deploying modern applications and need a high-performance network that also helps keep costs under control through simplified management. By providing any-to-any connectivity between switch ports, Virtual Chassis Fabric maximizes performance for physical servers, virtual servers, and storage in the data center. Virtual Chassis Fabric also supports VM mobility, ensuring that the network can keep up with the dynamic nature of today’s applications.


For more information see this paper, Juniper Networks Virtual Chassis Fabric Technology, and

Register for my webinar on May 21, Building the Right Switching Architecture for New Business-critical Applications.