Data Center Directions
, Regular Visitor
Data Center Directions
Choosing the Right Switch is Critical to your Transformation Project
Sep 29, 2013

If your organization is going through a data center transformation project you are probably looking at your options for switching infrastructure. As you design your network to support the move to virtualized compute infrastructure and the roll out of new application deployments the choice of switching infrastructure becomes a critical decision. One of the most versatile switching platforms on the market is Juniper’s QFX Series 10GbE/40GbE devices.  I’ve had the opportunity to talk with many of our customers about the projects that they are using the QFX switches for and why they chose it over the other options in the market. I’d like to share these examples with you.


Many organizations are undergoing data center transformation projects such as moving to a virtualized data center, deploying large scale Enterprise applications, converging data and SAN networks, and undertaking big data analytics projects. They are looking for a versatile switching platform that they can deploy in any of these scenarios. The QFX switches are high-performance, low-latency, 1RU edge devices that are installed at the top-of-rack in the data center. They include rich Layer 2 and Layer 3 support and standards-based bridging, routing and FCoE capabilities. They can be deployed as standalone switches and then as the deployments increase in scale they can be converted to a QFabric node through a simple software upgrade. This makes them ideal for these types of projects.


Server Consolidation and Virtualization
Organizations want to reduce data center infrastructure and operating overhead. They can do this by consolidating on large scale 10 GigE servers and running server virtualization software. With this move they can retire many smaller servers that were running one application per server, and run multiple workloads per server on the new infrastructure. They way they have fewer servers to deploy and manage and can optimize sever resource utilization across the infrastructure.

The challenge that customers face is the need for higher density connectivity to the network as they take advantage of virtualization and the high performance computer technology, and deploy more applications, and thus creating a proliferation of virtual servers. These new servers require multiple high bandwidth connections to handle the many virtualized workloads and the traffic between them. This move could be costly and it could introduce additional complexity so the right choice in network switching is important.

The QFX provides a high density of 10 GigE ports with low latency to insure performance for a high density of virtual machines each of which has virtual ports that need to be mapped to physical ports.  As the scale of deployments grows you can convert a QFX to a QFabric system. With the any-to-any connectivity of this architecture the IT staff doesn’t need to reserve adjacent racks for resources. They can allocate resources by TOR anywhere in the system, which simplifies the build out and prevents the need to deploy idle resources.  This means less energy and reduced OPEX and fewer spares to stock, reducing the burden on the operations and planning staff. To manage the physical to virtual environment from a single console they can use Juniper’s Network Director. For more on workload mobity see, Inter Data Center Workload Mobility with VMware.

Large Scale Enterprise Applications
Many organizations are deploying new large scale Enterprise applications such as Customer Resource Management or Enterprise Resource Planning systems. These applications are modular and distributed residing in multiple racks in the data center, and they create complex traffic patterns with considerable traffic running east-west between the components of the application.

Hops across the network bring increased latency that kills application performance, creating a poor user experience. This can be especially critical when these applications are being accessed from many locations such as branch offices and by supply chain partners. Large scale Enterprise applications require a network architecture that can handle the increased east-west traffic between the modules to ensure performance. This means that you need to reduce complexity in the network and simplify the architecture.

With rich Layer 2 and 3 capabilities at the edge traffic the QFX can optimized traffic flow and hair pinning can be eliminated. Low latency means that application traffic is not impacted as it moves between modules. A high density of 10 GigE ports provides high throughput. The Qfabric system provides a single hop, network that can act as a single large resource pool for complex applications. Any-to-any connectivity in this architecture means that nodes and storage can be placed anywhere in the system, simplifying the deployment process. For more see, Why Network Latency Matters for Virtualized Applications.

Data and Storage Network Convergence
Over time organizations built specialized networks for a variety of reasons. With larger organizations it is typical to find separate networks for data and for storage traffic. At one time this made sense as storage had special requirements that could not be met by Ethernet, however advances in technology have changed this.

Maintaining separate networks for storage and data is expensive and a management challenge. Many organizations are looking to reduce network complexity and eliminate redundant networks. One way to do this is to consolidate SAN storage on the data network using Fiber Channel over Ethernet. With FCoE you can eliminate the fiber channel adaptor in the server, and use only Ethernet, so you save on adaptors and cables and network management overhead.

The QFX supports I/O convergence, where it can act as an FCoE Transit Switch, and FCoE- FC GW, using standards-based capabilities that interoperate with your existing SAN. The QFX capabilities as an FCoE gateway reduces the number of gateways that are needed. It can load balance traffic for better performance and do FIP snooping for increased security. In the high speed QFabric architecture the QFX provides great flexibility in connecting to the SAN, and reduces the number hops in the network, improving FCoE performance. To learn more see, Making the Transition to Converged Storage and Data Networks.


Real Time Big Data Analytics
With the proliferation of diverse data sources, from social media, point of sale systems and machine sensors, there is an exponential growth in the amount of data that is being produced and there are considerable insights to be gained from analyzing it.  Organizations are seeking to find actionable insights from their data that they can use to drive the business in real time and increase their competitiveness. To meet this need they are deploying new infrastructure for running Big Data Analytics.

The challenge is that up to petabytes of data can be generated for many sources and it needs to be moved around in the network and operated upon efficiently during the analysis process. While Hadoop uses directly attached storage for the MapReduce processing, intermediate storage is used for staging the data. High throughput from staging to the Hadoop cluster is essential and low latency and low packet loss is critical when transferring data between clusters during the map and reduce operation to ensure seamless processing.

One design model for Big Data Analytics is to build on low end commodity hardware and let the software take care of everything. This model is used for non real time processing where completion time isn’t critical and data is primarily from one source. Another model, for doing near real time analysis on data from multiple sources, is to build a system upon hardware that provides deterministic performance to ensure continuous processing.


The QFX provides a high density of low latency 10 GigE ports that is suited for this environment. The QFabric architecture provides one large scale dedicated system for processing multiple large data streams with low loss, and with deterministic performance. This is critical for processing analytics workloads in near real time. Any-to-any connectivity in this architecture means that nodes and storage can be placed anywhere in the system, simplifying the deployment process. To read more see, How the QFabric System Enables a High-Performance, Scalable Big Data Infrastructure.

For More Information
With the flexibility to work as a standalone TOR or as part of a QFabric architecture and with rich Layer 2 and Layer 3 support, standards-based bridging, routing and FCoE capabilities, the QFX can be deployed for a variety of uses cases and provides a clear upgrade path as the scale of your deployments grow. It is this versatility that makes the QFX the switch to build your data center transformation projects on today and in the future. For more information see, QFX Series Switches on