If you’re a modern service provider you’re probably wondering: how to stay competitive in this rapidly evolving, technologically shifting industry? Answer: it’s about the network. Take a look at some key fabric technologies for your data centers!
Service providers come in different shapes and sizes. Sure, there are companies starting out in a new data center and just needing a couple of switches, but we also see many big players scaling up to hundreds or even thousands of devices, looking to build high speed multi-tenanted scalable IP-based networks.
One of the major challenges that they look to overcome is the manual configuration within the network. Obviously enough, that can lead to mistakes or inconsistencies, as they deploy technologies across the data center. One such example was Tussa IKT, a Norwegian service provider that supplies the residents and businesses of Sunnmøre district with the Internet, TV and advanced IT services.
“It was becoming increasingly challenging to conduct regular maintenance work. Daily operations such as establishing, configuring, verifying and troubleshooting VLAN were complicated, due to many switch rows that had to be changed manually”, says Ivar Driveklepp, CEO of Tussa IKT.
Tussa IKT was convinced that replacing the old core network would provide a much better way to operate their large networks. And so, a call for tender was issued.
“At the beginning, we were expecting to go for MPLS to meet the future challenges. It was not until the tender round we heard about Extreme Fabric Connect”, Ivar Driveklepp adds.
In general, many service providers build a small network initially, and then they need to grow that by adding number of switches in the future. With that in mind we came up with Extreme Fabric Connect – a Shortest Path Bridging (SPB) solution that helps you manage the configuration consistently across the network and ensure the network scales in an even way.
No more manual configurations, no more errors
Whether it’s just a couple or thousands of switches, Extreme Fabric Connect allows to deploy configuration in a single, consistent manner across the whole network, providing a platform for the service providers to deploy new tenants or services. The main operational benefit many of our customers see is that it removes the manual configuration or changes on the network.
That means no more errors or inconsistencies in the way that configuration is applied within the network. End-to-end services, including IP multicast, are only provisioned at the edge of the network. Once the services are provisioned at the edge, the fabric automatically connects the services across the fabric, eliminating risky and time-consuming hop by hop provisioning practices.
Wish to go in more detail? In the context of Data Center networking, Extreme Fabric Connect…
- …simplifies virtual machine mobility by stretching IP subnets or Layer 2 domains within and between data centers maintaining the IP address of the virtual machine regardless of its physical location. These stretched VLANs, referred to as Layer 2 Virtual Service Networks (VSNs), can extend anywhere in the Fabric Connect network and can be set up with ease at the network edge.
- …simplifies Data Center Interconnect by enabling active/active, full mesh connectivity between multiple data center locations while ensuring a logical loop free environment without any additional overlay technologies.
- …offers dynamic auto-attach features for ESX-, HyperV- and KVM-based hypervisor environments so that as virtual machines are spun up and moved, network services are extended dynamically.
- …simplifies both multi-tenant/micro-segmented networks through integrated VRF capabilities, referred to as Layer 3 Virtual Service Networks (VSNs), which can set up with ease, without any additional overlay routing protocols.
- …can be deployed over any physical topology whether it be traditional spine/leaf, mesh, partial mesh or other topologies – with path optimization for all traffic flows.
And how did the aforementioned Tussa IKT benefit from introducing Extreme Fabric Connect, specifically?
“Easier maintenance work and administration, greater opportunities for automatic redundancy through mesh and ring structures, less need for physical changes, fewer error situations, establishing services is much less time-consuming and can be done with less risk…”, Ivar Driveklepp can’t stop counting the benefits.
Extreme Fabric Connect… and what else?
Once an SPB core is in place, service providers can deploy additional features to increase the reach and the value of the deployment:
- Dynamic Auto-Attach: the functionality provides for automatic attachment of users, devices, and virtual machines (VMs) to connect to SPB services or I-SIDs. It uses extensions to the IEEE802.1AB Link Layer Discovery Protocol (LLDP) to automatically attach network devices to I-SIDs or VSNs in an SPB network. This auto-attach capability can be deployed on endpoints, such as IP surveillance cameras, wireless APs, and/or non-SPB-compliant access layer switches so that seamless communication with the SPB fabric is possible.
- Seamless extension of SPB: Fabric Extend is a feature that enables SPB to be extended over third-party IP networks whether they are IP cores or service provider wide area networks (WANs). Any SPB service, whether it is L2, L3, or multicast, can be seamlessly extended across the WAN or across the IP core easily — without the WAN or IP core having visibility to those services. Fabric Extend supports an IP mode, an L2 mode or an encrypted mode. It works by creating an overlay of point-to-point tunnels using either Virtual Extensible Local Area Network (VXLAN) native L2 tunnels for IP mode or encrypted IPSec tunnels.
Distributed Virtual Routing (DvR)
In addition to achieving significant organic growth, many service providers also grow inorganically through mergers and acquisitions. This often poses a challenge of integrating “inherited”, traditional networking technologies into a single end-to-end fabric. To further optimize traffic forwarding and enhance resiliency, Extreme Fabric Connect comes with another SPB feature called Distributed Virtual Routing (DvR).
In short, DvR improves the scale and performance of multi-site networks with mobile end-points. These can be either VMs or wireless clients at the network edge. The main value of the feature is that it eliminates the “tromboning” problem that can occur when the hosts are too far away from their default gateway due to stretched VLANs that might span a Data Center or even multiple Data Centers.
Where routing is centralized traffic will have to traverse across the network, even though the host source and destination of the flow is local to the rack.
DvR distributes the routing function to all switches that have a presence in the IP subnet so that the default gateway is always available at the first network hop. At the same time, it keeps provisioning simple because the routing configuration is performed only on centralized controllers or spine/ aggregation nodes.
Distributed Virtual Routing distributes the routing function without having to provision any routing functionality on the ToR or leaf nodes. The result is optimized traffic without complexity.
What’s key with DVR is that even though the routing is distributed, the provisioning is centralized and conducted only on designated aggregation / core switches. Therefore, you can take advantage of the performance of distributed routing while keeping the network streamlined and simple by centralizing the provisioning.
If you’re looking to better optimize routing, eliminate traffic tromboning and reduce latency within Data Centers that are leveraging virtual machine mobility, you should look at Distributed Virtual Routing. DvR can replace VRRP in the Data Center, providing a simpler, more scalable, and higher performance solution.
Wish to discover more Extreme Fabric Connects capabilities and see how they can benefit your organization?
Tussa IKT - read the full customer story
Tussa IKT: experiencing the difference between previous generation network and Fabric Connect first-hand
Working with IT networks could be considered an art, rather than an actual field of expertise or subject taught in a technical school – just ask any network administrator! Complicated setups, high risk of running into costly mistakes, delicate configurations… it all requ...
Read More