Data centers are the heart of the cloud that everyone has come to rely on, but what is the role of data center interconnect and why is it important? I recently met with Mervyn Kelly, EMEA Marketing Director, and during our chat he offered some fresh perspectives on the role and value of the communications network that connects datacenters together.
SF: So tell me, what is DCI?
MK: DCI stands for data center interconnect. It is the network that connects datacenters together and enables and underpins any cloud deployment – private, hybrid and public. DCI is one of the key building blocks of cloud networking and content distribution.
SF: How are data centers connected?
MK: Users connect to the data center using the Internet, IP VPN connectivity, and increasingly via Ethernet business services to provide high quality high capacity connectivity. But, to connect data centers together, high bandwidth low latency Ethernet and optical connectivity are needed. The optical networks can be self-built with dark fiber, or a leased circuit from a communications service provider, or some combination of these.
SF: Seems simple. A data center owner either buys or builds an optical network. Is it really that simple?
MK: Well, it’s not as simple as ordering a mobile phone. One of the biggest problems is the time it takes for a provider to turn up a wide area network (WAN) circuit. Long lead times for an enterprise are like long queues at the market, no one is happy. Haven’t we all walked away from a long queue, leaving our basket behind? Just as an abandoned basket impacts the market owner’s revenue, a data center operator that projects $2.4m revenue per year, will be impacted $200k for each month that a WAN service request is delayed.
SF: I get it. I have abandoned a purchase waiting on a long queue or for a web page to load. It makes total sense that speed of deployment would be a key driver. Do you have other examples of challenges?
MK: The majority of DCs are now in metro areas to be closer to their customers. London, for example has over 337 DCs. This multitude of facilities drives other pain points such as space limitations, power cost and availability, ability to scale, complex operations and dark fiber availability. Oh and cost. The more technical fingers required to operate the data center, the higher the cost and the increase in the chance for human error. So operators are looking at automation and plug and play functionality as ways to reduce costs, accelerate service turn up and add new services.
SF: That’s a lot of pain for a data center operator. What about the needs of the enterprise end-user?
MK: As bandwidth continues to grow, the WAN network needs to have the appropriate bandwidth and performance characteristics to suit the service.
For example, data backup, between primary and secondary data centers, can be extremely slow on low bandwidth connections, and for real time synchronous backup this will have a big impact on the application performance. One terabit (1 TB) of daily data backup between DCs over a 100Mb IP MPLS circuit will take 1 day. By the time you finish the backup it will already be out of date! The same 1TB over a 10Gb connection will take 13 minutes, or over a 100G connection will take 80 seconds. With business critical data, which would you prefer?
SF: What are the specific DCI services that a provider can offer?
MK: Here are a few:
- Backup – is the most complex of all IT operations, touching every part of the infrastructure. Backups are the bedrock of an organization’s compliance activities, and the basis for archiving and vaulting.
- Business continuity and Disaster Recovery (BC/DR) - comes from being able to restore services to alternate locations – fast RTO/RPO’s (recovery time objective, recovery point objective) are business critical.
- Hybrid cloud connectivity - relies on being able to access and move large amounts of information between interconnecting private and public cloud data centres, often with varying protocols.
- Algorithmic trading - with computer based stock trading, time really is money and high speed is required with the lowest latency, to enable optimal execution of trades.
- Load balancing and virtual machine (VM) migration - moving workloads between servers to ensure optimal application performance requires short- term high bandwidth. It has been recommended that a minimum of 1Gb/s of bandwidth be allocated for each VM move.
- Content distribution – OTT players such as Netflix need the ability to distribute their most frequently watched content between centralized and distributed DCs not just with high bandwidth and low latency but to different locations at different times.
SF: Do you see new applications coming along enabled by new technologies?
MK: Yes, the next logical step is to truly align the WAN DCI with the application, and deliver SDN (Software Defined Networking) enabled on-demand connectivity services that are paid for based on usage.
SF: Speaking of SDN, what is the role of virtualization with DC applications?
MK: There is definitely a new market opportunity for DC operators to offer virtualised IT applications to enterprises. Virtualized firewalls and virtualised routers are two applications that are top of mind. Since NFV (network function virtualization) is based on software, the cost of entry into this market, both CAPEX and OPEX (News - Alert), is low and can be success based – there is no hardware to buy and deploy up front other than standard servers, which may already be in the network. An operator can pick the best of breed VNFs (virtual network functions) in the market, and easily change VNF vendor if required.
SF: We’ve covered a lot of ground in this interview. Can you summarize our conversation in a nutshell?
MK: Certainly. Users, enterprises and consumers, require connectivity to data housed in a variety of DCs. To deliver high performance, the connectivity between these DCs needs to change. DCI that is optimized for the modern DC needs to have the attributes of high scalability, openness and programmability, simple operations, minimal space requirements and low power usage.
And while not strictly part of the DCI solution, there is a new revenue opportunity for DC operators to deliver virtualized IT applications, using NFV and an orchestration platform to deliver and manage the VNFs.
Edited by Maurice Nagle