2015 was a seminal year for data center interconnect (DCI) technology. It was the year that ever increasing spending by big data center operators and Internet content providers finally made the communications industry take note. No longer second fiddle to large Tier 1 Service Providers, large data centers were finally heard when they said “we need something different”. And the optical communications industry responded in a big way, not just tweaking existing products, but launching entirely new product lines to serve their big needs (and big spend). But if DCI hype reigned supreme for 2015, what does 2016 have in store for us all? Well, with the disclaimer that one predicts the future at his own peril, there are 5 major trends that I think show signs of being huge for DCI in 2016:
1. Power Efficiency Will Continue to Dominate
Power efficiency has always been important to DCI and it will stay on the agenda in 2016. Data center operators have plenty of money and space, just not power. They are literally building underfilled warehouses in the hopes that future power efficiency gains will allow them to fill them to capacity.
2. Peering Will Be a Key Growth Area
Peering is basically the direct sharing of customers between content providers. It’s very inefficient to go to the public Internet and back again when the same customer is accessing servers in a common colocation facility. At the minimum, it’s a cable between two adjacent servers. At the maximum, it may be a dedicated transport link between cities. But no matter the physical link, only a limited amount of customer parameters and traffic is shared, and never are two competing private networks physically connected. Click here for more information on the types of peering.
3. We’ll See a Shift Towards Physical Layer Encryption
The entire Internet uses IPsec for security. So when it was discovered that IPsec links are relatively easy to compromise, data center operators and internet content providers turned to encryption higher up in the Open Systems Interconnection networking stack, namely Secure Sockets Layer (SSL) and custom variants of it for enhanced protection. They are now realizing, however, that the same techniques used to compromise IPsec work for SSL as well, and the only true way of protecting data as it transverses between their secure data centers is physical layer encryption on the fiber itself.
4. Open Architectures Will Appear
Openness will transition from theory to practice in 2016, on both the hardware and software sides.
On the hardware front, Open Optical Line Systems will appear that allow data center operators to leverage the fact that while servers and switches may be replaced every three years, the optical line system gear-like channel multiplexers and optical amplifiers can last much longer, and fiber is forever. Data center operators will now be able to build fiber optic information highways that any server, switch, or router can ride between destinations.
On the software front, there are two important open efforts underway, one driven by service providers, and the other by content providers. Service providers as well as other industry leaders are working within the Internet Engineering Task Force to standardize on a common open software approach with its roots in OpenFlow. Meanwhile, several data center operators and content providers are putting their support behind OpenConfig. Both will enable future transport networks to be built that can be fully controlled by a top-level orchestration layer that is able to optimize multilayer end-to end packet flows.
5. The Trend to Flat Network Architectures Will Impact DCI
As consumer and business usage trends have changed, so has the traffic within networks. Today’s traffic patterns have created a rapid growth in “horizontal” traffic, which is data passed from machine-to-machine within a network. “Vertical” architectures that rely on pyramid like hierarchical layering are inefficient when processing horizontal traffic. Thus, data center operators and Internet content providers are looking to remove the vertical layers of their network. For example one major operator would like to move from five vertical layers to only two. This paradigm shift in network architectures is having a profound impact on DCI as well. In order to support these new flat network architectures, transport gear must provide much higher client port counts, channel density, and fiber spectral efficiency.