The application architectures have evolved from client-server to distributed apps, causing an underlining shift in traffic pattern. The newer applications are increasing the communication between server to server causing more east west traffic within the data center.Server virtualization is helping businesses gain efficiency by consolidating many physical servers into fewer high-performance virtualized servers. Of all the traffic types traversing an Ethernet network, storage has risen in prominence in recent years. This traffic type has its very own unique characteristics and demands a separate treatment than other traditional Ethernet traffic.Networking has not progressed and is now a barrier slowing the pace of innovation in the rest of the data center.
- High performance: eliminate STP, 1G->10G in access and 40G Fabrics Scale: seamlessly scale from 10s to 1000s ports without disruptions and Operational simplicity by reducing the managed devices and total cost of ownership
Concept of a chassis switch: We have line cards, which contain the ports. We have some sort of a center plane or back plane that you plug into. Then there is fabric circuitry, which interconnects all the ports in each line card to all the other ports in the line cards. Ethernet packets come in, some initial processing at the egress port, and then the bits are sprayed across the backplane in a non-blocking fashion and reassembled at the far side and then Ethernet out. What you experience as a user is “Ethernet in and Ethernet out”, but you don’t manage these bits. It’s a very efficient transport. QFabric also behaves like any classis switch. QFabric has distributed components but is managed as a single logical switch. In designing QFabric, Juniper has essentially taken the three basic components of a self-contained switch fabric—line cards, backplane, and Routing Engines—and broken them out into independent, standalone devices—the QF/Node, the QF/Interconnect, and the QF/Director respectively. • QF/Node: In the QFabric architecture, the line cards that typically reside within a chassis switch become a high density, fixed configuration 1 RU edge device called the QF/Node, which provides access into and out of the fabric. The first member of the QFX Series product family, the QFX3500, not only acts as a QF/Node edge device in a QFabric architecture, it can also serve as a high-performance standalone switch in highly demanding data center environments. • QF/Interconnect: Similarly, the backplane of a single switch becomes the QFabric/Interconnect device, which connects all QF/Node edge devices in a full mesh topology. • QF/Director: The Routing Engines embedded within a switch are externalized in the QFabric architecture and called the QF/Director, which provides the control and management services for
Simulating data traffic representative of today's most demanding data center and cloud environments.The QFabric system showed that enterprises can scale their existing data center infrastructure without loss of performance and without adding complexity. The QFabric system demonstrated record performance, delivering network multicast traffic at a rate of 15.3 terabits per second -- enough bandwidth to stream 3.4 million HD movies simultaneously.Results confirmed that all 1,536 ports were easily and simply managed as one device.The testing also validated interoperability with a variety of switches.
A change in data center network design is needed to ensure that organizations can take full advantage of their investments in new applications, virtualization, and storage and compute resources. The most efficient way for resources to interact is for them to be no more than a single hop away from each other. It’s time to break the network barriers and build a network environment that is optimized for performance and simple to operate.QFabric architecture would address the latency requirements of today’s applications, eliminate the complexity of legacy hierarchical architectures, scale elegantly, and support virtualization, convergence, and cloud computing and other demanding requirements for the next-generation data center.
A Revolutionary New Switch 3 Design PrinciplesManagement N=1 Operational model of Plane a single switchControl Plane Federated Intelligence Only way to scale with resilience Data Plane Rich edge, Simple core Everything is one hop away