We believe that the network is keeping the business alive. It is the heart of the business, and it is the most sensitive aspect of any business. Downtime is not acceptable, old slow converging protocols are frowned upon in the industry. Reliability, redundancy and 100% uptime is the new norm. Without it, the business is dead.

We were tasked with the complete upgrade for a webhosting company which experienced huge success with their service offering and were in great demand of a complete re-design. They were providing services for more than 50000 users which were very sensitive to the smallest latency increase. And rightly so.

Basic Concepts

- Existing Network build across multiple locations in the UK; - A mixture of vendors – Cisco and Juniper switches;
- Dark Fiber links to interconnect the locations;
- The network was built around switching protocols with Spanning Tree in the core;
- No support for IPv6 or MPLS;
- No traffic load-balancing across multiple paths.

Project Requirements

- Upgrade the network to Juniper only;
- Deploy DWDM on the dark fiber to increase capacity;
- Enable IPv6 support;
- Enable MPLS and VPLS technologies;
- Remove Spanning Tree from the core;
- Implement a layer 3 only core, with multi-homed BGP upstream providers;
- Migrate the existing services to the new network without downtime;
- Enable Multi-Site Redundancy using redundant VPLS circuits.

Design and Testing

Phase 1

- Deploy wave division multiplexing to allow additional capacity for the deployment and testing of the new network in lab environment;
- Install 64 DWDM Channels across the fiber links, with multiplexers, media converters and amplifiers.

Phase 2

- Build the new network out of service using the additional capacity provided by DWDM, as per the requirements listed above;
- Test routing scenarios and failover scenarios. The decision was made to use OSPF inside the network and BGP for the multi-homed upstream providers;
- The limited power rackspace available required careful consideration for the vendor equipment that would support all the features needed. It was decided that Juniper MX series best fit our needs and as such testing started for the suitability;
- Two months and three software versions later the new network was ready to be deployed in service, using a tested and stable software which provided under 20 miliseconds failover on any of the links.

Phase 3

- Connect the new network to the existing switches;
- Migrate the services to the new network without downtime;
- Retain part of the old network (mainly the edge switches where all the servers were connected to);
- Decommission the old network devices which became unused as a result;
- Hand over the completed project four months later.

Key Advantages

- The customer now benefits from new technologies like IPv6, MPLS and VPLS. Another key aspect is the improved failover which is totally transparent for the services provided by the customer, in case of a link failure. Traffic is re-routed around under 20 miliseconds, resulting in fewer tickets for the customer support to handle, no downtime, no impact on the business;
- Wave division multiplexing now enabled the customer to deploy up to 640 Gbps capacity. This allows for great scalability at a very low CAPEX cost.