Industry Outlook is a regular Data Center Journal Q&A series that presents expert views on market trends, technologies and other issues relevant to data centers and IT.
This week, Industry Outlook asks InterOptic CEO Tim Dixon about optics and its increasing role in IT. InterOptic is a data-interconnect company that helps Fortune 500 and government institutions manage the bandwidth, interoperability and complexity of their IT networks. Employing decades of optical expertise, the company offers a range of data-interconnect solutions, including a full line of industry-standard optics hardware that’s identical to the technology OEMs use but at a fraction of the cost.
Industry Outlook: What exactly is optics?
Tim Dixon: In the context of the data center, optics is short for the fiber-optic cable and laser transmitters/receivers that interconnect most data paths between networking, compute and storage elements.
IO: How do data centers benefit from integrating optics technology in their designs?
TD: The use of lasers for data transport dates back to the 1970s, but in the early 2000s we saw the rapid adoption of pluggable optics that provide flexibility in deploying data rates and overall bandwidth, along with easier equipment upgrades and maintenance. This development, coupled with advances that reduced the size, cost and power consumption of the technology, are all benefiting the data center today. Companies that invest in optical networking hardware and design (e.g., optical transceivers) will increase reliability, speed and interoperability across their IT networks to meet increasing bandwidth and scaling requirements.
IO: What technologies and trends are stretching the limits of data center networking and what’s the result?
TD: Optical signal to noise imposes a limit that caps the aggregate data that can be transported over a single fiber. We’ve been testing this limit for almost a decade, first in long-haul (inter–data center) and now short-range (intra–data center) connections with speeds of 200Gbps and above to accommodate the appetite for 5G, IoT, big data, quantum computing, and other emerging technologies. This demand has increased the need for speed, uptime, reliability and efficient data transfer. In the data center, it’s led to the proliferation of parallel optical paths that must be terminated, monitored, managed and processed on each piece of equipment. The associated costs and complexity in the electrical realm are increasing; future data center design begs for simple, low-cost ways to distribute some of this complexity into the optical layer for overall greater throughput.
IO: Light is light, so what factors determine which optics solution or project is best?
TD: Yes, light is light, but just as you would see a blurry reflection in a tarnished mirror, the elements of producing and transferring that light have a great impact on its integrity. The design-element selection plus the testing, monitoring and supply-chain traceability all work together to provide a high-performing optical component that will last for decades in the field.
IO: Given that bandwidth requirements will only increase, what’s the future of data interconnects?
TD: This question relates to the “stretching of limits” in the data center. In many ways the data center of today has inherited the siloed approach to equipment design typical of computer and networking architecture, along with the hierarchical bias of the OSI model. This segmented, box-like mentality is a classic systems approach, but it isn’t always the most efficient, most creative and most cost effective when you’re looking at the big picture. I believe you need a more holistic approach to tackle the data center of the future—one that converges logical and physical boundaries we see today; optics of all kinds will play a role.
IO: How can enterprise companies solve this problem of quickly and affordably increasing their data center bandwidth?
TD: If you look at the near term, don’t paint yourself into a corner. We see a trend toward copper interconnects. These solutions, although cheap today, hit a brick wall in overall throughput as well as distance and route-diversity limits. We’re surprised at how few VPs and CIOs look at the life-cycle cost of hardware in their data centers. Making correct choices on this “boring stuff” can save millions and boost the success of other critical IT initiatives. Consider looking into brand-equivalent hardware providers, which can outperform OEMs such as Cisco and HP at a fraction of the cost.
IO: How does optics affect the cloud?
TD: “The cloud” as we view it is nothing more than where compute and storage resources are deployed. The infrastructure in any cloud data center looks strikingly similar to that of any Fortune 1000 internal data center.
IO: Is there a “right time” to replace optical transceivers in a network environment?
TD: With our customers, we don’t say “replace”; we say “redeploy.” The longevity and reliability of our products should mean you never have to replace them. We have large federal-government deployments whose optics have been in the field for close to a decade. They’re just unplugged and used farther out in the network where the bandwidth needs are less and are a great match for those earlier-generation optical transceivers. Then, as the network and data center core deploys new and higher-performance architecture elements, they can move to higher-throughput and higher-performance transceivers.
IO: What are some of the biggest trends you’re seeing in large data centers?
TD: One of the biggest trends we see is the movement away from large data centers. Sure, you’ll still have the five or six big ones, but many industry segments are moving to micro data centers. This is especially true when you have real-time, low-latency applications such as gaming and high-frequency trading. As mentioned earlier, a more holistic approach to design should prevail within the next decade: data centers won’t be “one size fits all” or “one approach rules,” but instead customized and targeted per industry and application.
IO: What are some common misconceptions about optics that you’d like to dispel?
TD: The myth that all optics are the same—this idea couldn’t be further from the truth. Every member of our leadership team has been involved in optical networking for 20+ years. When we were designing early systems and solutions at Bell Labs, AT&T, Rockwell and so on, we certainly made choices and tradeoffs that mattered. Even today, selecting components, analyzing and testing critical parameters, and how we do our coding all make a huge difference in the performance and reliability of our data centers and networks—and therefore in customer satisfaction. Few (if any) OEMs provide this level of expertise, persistence and personalization when it comes to designing and deploying optical hardware.