The rapid growth in IT infrastructure, including business continuity applications, cluster deployments, and server consolidation across the enterprise, has resulted in a bandwidth explosion inside the data centre with 10G high speed ports becoming more common on datacenter routers, switches and server clusters.

At the same time IT managers are faced with requirements to ensure data consistency, increased redundancy and rapid failure recovery across datacenters driven by new regulations in HIPPA, Sarbanes Oxley and recognition that highly centralized datacenter applications expose businesses to unnecessary risk.

The challenge that many IT managers have come to expect in this environment is the difficulty and expense in securing bandwidth capacity between data centers. Often the high cost causes IT managers to re-design architectures between datacenters to save cost and utilize schemes that can be delivered over low rate T1 circuits when leaving the building.

This mismatch between data center bandwidth and the need for cost effective high speed interconnect to manage the business can now be addressed economically with recent changes in the fiber optic system transport market that bring high speed connectivity within reach of more typical low rate interconnect options.

With the increased availability of fiber in the metro and the emergence of a new class of fiber transport products - thin WDM - IT manages can now exploit large bandwidth interconnect for a fraction of what was possible only 4 years ago. Thin WDM systems provide cost efficient wavelength multiplexing in small 2U fiber transport systems delivering 10G port extension and 10G Ethernet private line between datacenters for a fraction of traditional WDM systems.

Enterprise managers can now take advantage of the value of direct wavelength connectivity between datacenters at rates up to n*10G for reliable, redundant, high bandwidth at price points that enable building distributed architectures between data centers. With 10G interconnects, IT managers can look at scaling for rapid recovery from failures, timely backups and restore, and effective cluster applications between data centres across campus and metropolitan area networks.

Why now?

Three key technology and market advancements are driving down the cost of connectivity, dramatically enabling IT managers to design more distributed data center architectures.

Fiber availability: The last six years has seen the emergence of metro fiber from many alternative sources. i) Utility companies now provide fiber paths along their rights of way, ii) a new class of CLEC's has now consolidated fiber capacity in metro areas that is now broadly available, iii) city and municipal governments have installed dark fiber and iv) incumbents are now rolling out building fiber assets further out in their networks passing key business sites. For IT managers this means that the probability of being within 100 feet of fiber is higher than before.

Fiber interface advancements: With the emergence of CWDM (Coarse Wavelength Division Multiplexing) and cost effective SFP's (Small Form Pluggable optics) it is now possible to light a fiber for under $2,000. Many of today's servers, switches and routers now support standard SFP's making it easy for them to network directly on fiber over short distances. Standard, volume interfaces are now becoming widely available making interconnection with wavelengths more feasible than ever.

New class of fiber transport platforms: Fiber transport systems have emerged that can prove in on the first channel (wavelength) and grow very efficiently to serve 8 or up to 32 channels of capacity on the same fiber. These systems typically prove in on the first 1 or 2 interfaces and then provide "free" capacity for subsequent channels with Wavelength Multiplexing. With cost effective wavelength multiplexing of these systems coupled with standard wavelength interfaces on routers, switches and servers, these fiber transport systems provides high bandwidth interconnect with performance monitoring metrics ensuring bandwidth outside the datacenter as inside.

10G networking alternatives outside the data center

Cost effective 10G on fiber transport can be delivered typically in two ways.

The first option to consider if dark fiber connectivity can be secured (leased or owned) between data centers is setting up a private network utilizing the massive bandwidth capabilities of fiber coupled with the effective interconnect with a Thin WDM system to manage the fiber capacity and interfacing of various signals onto the fiber. In cases where dark fiber can be secured, deploying Thin-WDM provides cost effective channel "light up" that also provides "free" wavelength growth via addition of wavelengths or colors as the application environment grows.

The second alternative is to see if a progressive service provider in your area is willing to build a managed wavelength service. Often you will find a utility company, or CLEC that has the ability to serve you quickly and is willing to move beyond the large core network 32 channel systems that large carriers force fit for smaller networks. This option involves purchasing a managed wavelength service with a Service Level Agreement from a service provider typically charged on a per wavelength per month basis. Where a managed wavelength can be purchased it can be specified that the service must be delivered on Thin-WDM platforms to ensure you are getting a solution that is economical for both you and your service providers for lower wavelength counts.

Fiber transport platform considerations

When selecting a platform either for a private build on dark fiber or to provide a managed wavelength service from a carrier IT managers should be careful to specify a number of items to ensure flexibility and smooth growth options including:

Port interfaces should be SFP based or some form of modular interface that defers the laser cost until you light the channel
The platform should be able to provide service demarcation of standard 1310, 850 and wideband interfaces already existing in the data center - this ensures seamless interconnect and avoids a forced router, server or switch line interface upgrades..
Reach capabilities - depending on the path of the fiber, straight line distance versus actual fiber distance can be 2-3X so it is worthwhile to ensure any platform you select has slide-in optical amplifier cards when fiber performance or distances get further than the standard 40km - with this capability you can easily extend your planning distance to 100km and beyond giving flexibility in network design.
CWDM and DWDM capable - many CWDM platforms are limited to 8 channels and require a forklift upgrade to go beyond 8 channels to DWDM which scales to 32 channels - newer systems can handle both CWDM and DWDM seamlessly transparently to the user giving the best of both worlds - startup costs of CWDM and the reach distances of DWDM.
Modular - ability to upgrade to 4G FC and 10G easily with no disruption
Fiber protection - does the platform provide fiber protection and sub 50 ms optical path switching?
Footprint and power - look for 2U systems that can be installed and in-service in 30 minutes with minimal training - newer systems have a much more intuitive setup and installation model than older systems designed in the 90's which will dramatically reduce support, installation and upgrade costs.

With these things in mind IT managers can now think more freely about the old adage "if bandwidth were free, how would I design my network." While not quite free, the new capabilities now on the market provide a compelling shift in bandwidth costs that can dramatically change how you build and grow your enterprise network providing improved connectivity between key datacenters. ENS

Glenn Thurston is vice president of Marketing at BTI Photonic Systems, a company that provides Thin-WDM solutions to enterprises, utilities and carriers introducing cost effective wavelength transport that changes the game for how wavelength services can be delivered. He can be contacted via www.btiphotonics.com/ company/ team.php.