The nirvana of a data network infrastructure is something akin to light bulbs: When you want light, you switch it on. When you don’t need it anymore, you switch it off. If you go into a different room, you switch on the light in that room then turn it off when you leave. Or, if you want light but not too bright, you can dim the bulb. And of course, you only pay for the power you use while the lights are on.

Similarly, in the world of data networks, firms want to be able to switch services and routes on and off, on demand and as needed without financial penalties, and to be able to increase bandwidth when needed and dial it back when not needed—like dimming a light bulb. And increasingly, firms want to be the ones flipping the switch.

These networks exist today. Call them client-provisioned or client-configured, or—to use the technical name—software-defined networks, they allow staff at a network provider or a client firm to dynamically configure networks in real time. 

“Self-provisioning is very much a natural evolution for networks,” says one bank CIO. “It’s kind of a dream come true: imagine you want to increase bandwidth—from a customer perspective, it’s very good.”

Scott Caudell, CEO of FiNext Networks, a cloud technology provider to financial firms, goes further, describing a “chasm” of demand from firms looking for more control over network resources yet still trying to compress IT costs to free up budget in other areas.

“From a customer perspective… what has worked for them over the past six months may not work over the next six months, so firms need to be able to change plans, start new initiatives. And the interval between needing to start new strategies is now shorter. It used to be that if you found a source of alpha, you had a six-month to two-year advantage,” Caudell adds. 

Agile and Flexible

One reason SDNs’ supporters like the concept is that being able to change network configurations themselves without involving third parties allows them to be more agile and take advantage of opportunities as they present themselves. For example, if a firm wants to take advantage of arbitrage opportunities in global foreign exchange markets, or believes it has found a source of alpha in an asset class and wants to test its hypothesis, or dip its toe into a new regional market without making a large investment in overseas infrastructure, it can do so, assuming its network provider supports it.

“Flexibility is the big driver, and it’s hard to be flexible with the old static networks, which can take days or weeks to get something done, rather than instantly. SDNs offer cost efficiencies, but also revenue benefits, because you can do more in a shorter period of time,” says Guy Warren, CEO of systems monitoring technology vendor ITRS. 

“Clients can switch services on and off, and can pay for what they use, on demand, as and when they need it, and not pay for what they don’t use,” says Rob Coole, director of product management at IPC Network Services, which will roll out its own SDN offering late this year or early next year, having already rolled out a self-provisioning client portal for its voice networks around three years ago.

“When firms see that they can replace physical assets, and take that budget and move it around while achieving the same level of alpha, that’s a no-brainer to me. And your downside in this new world is so diminished that if something doesn’t work out, who cares? The costs we’re talking about are chump change to these firms,” Caudell says.

SDNs also offer benefits to the network providers themselves. “The benefits for IPC are around service delivery. The sooner we can deliver a service for customers, the sooner we can start earning revenue. Instead of having to go to all our providers and wait four to six weeks for our underlying carriers to enable a new port… we can provide a service and start billing right away,” Coole says.

Time to deployment is the key issue for providers, according to Caudell. “Say it took six weeks to deploy a service to a customer… now, you can take that from six weeks to six minutes. You recapture six weeks of revenue. Your risks and labor costs go down substantially, and you gain more consistency and customer satisfaction when you’re done.”

In addition, third parties, such as financial technology startups seeking channels to market and established fintech providers, can also benefit from SDN capabilities.

For example, vendors like broker-neutral front-end trading platform provider Sterling Trading Technology, which need to provide trading connectivity to clients, could benefit from being able to switch routes on and off depending on client demand, rather than maintaining expensive links full time. Sterling likes to build proprietary connectivity where possible so it can provision clients immediately, but also coordinates services from third-party network carriers.

“You want to provide a one-stop shop for clients so that when you get a requirement, you can respond to it quickly… and they can execute on their plan,” says Andrew Actman, director of business development at Sterling. “Everybody is looking for an edge—a new market, a new execution route, or a new asset class… and the last thing these customers want is to have to worry about all those things that are not core to what they do.”

Though Sterling’s setup doesn’t constitute an SDN, it could clearly benefit from its underlying carriers being able to offer the service. Similarly, because IPC builds its network on physical infrastructure operated by third-party suppliers, it is also a customer in the same sense as its own clients—and the ability to self-provision in real time is as important to the vendor as to its clients. In fact, IPC refers to itself as an “orchestrator” of third-party networks and APIs. 

“We’re leveraging their APIs that control the status of their networks in real time. We need to be able to query that carrier’s network… to show routers, bandwidth utilization over each link, and so on,” Coole says. “If we weren’t able to query those networks when accepting an order, it would delay the process—and that wouldn’t be a customer portal; it would just be an order form.”

It’s a similar story at UK-based network provider Colt, which has invested heavily in its network of late—specifically in this space, which “empowers users to take control of their cost base more efficiently,” says portfolio director Tim Williams. He says Colt can offer SDN capabilities in Europe and Asia, and is now building out that offering in the US.

Connection Challenges

However, there are some practical challenges that must be overcome before the industry as a whole can take advantage of self-provisioning.

One challenge is technical: to be able to self-provision a route to a particular marketplace or building, that route must already be on net.

“For a network with plenty of connectivity already in place, allowing clients to self-provision is easy. But connecting to new buildings can be a lengthy process,” the CIO says.

Colt’s client portal allows users to type specific addresses to see if a building is already on Colt’s network, which comprises connectivity to 800 datacenters—currently six in the US, though the vendor plans to grow this to 30 by year-end—and 25,000 buildings. 

“If a site is already lit and has equipment installed, we can be done within minutes. The longest delay is usually the time it takes a [new] customer to plug in to our network. And if it is not on net, we would have to run fiber into the building, or maybe use someone else for the ‘last mile,’” Williams says.

Another challenge is more an issue of perception and firms being comfortable with the idea of new technologies—similar to the way that the use of public cloud computing and storage services has in the past couple of years rapidly gone from being something that struck fear and uncertainty into the hearts of capital markets IT executives to something that is de rigeur.

“Retail markets are already used to the capabilities of the internet, whereas capital markets still prefer private communications lines. The majority of investors are benefitting from investment in the internet rather than private comms: bandwidth is increasing, and the price of bandwidth is falling, while the security of the internet is also improving,” says the bank CIO, whose role covers technology serving both the retail and investment banking divisions of his firm. “But most of the networking out there still uses legacy technology that doesn’t allow for client configuration—and it will take a long time to upgrade and replace those networks.”

While Colt, IPC and some other network providers are already well down the path to self-provision, one senior network professional at a securities broker warns that the cost firms will ultimately pay for the privilege of self-configuration will depend on where a supplier is in its lifecycle, and that promises can often exceed reality—for example, not just around being able to deliver on-demand configuration, but being able to properly implement the associated on-demand billing features. 

“Networks involve a lot of very physical components. You may be able to circumvent the order process, but a lot of components and automation features around that still need to be built by network providers—and they aren’t there yet, and I think it will take a long time before many of them can become customer-configurable,” the network professional says, adding that in the meantime—and especially in instances where a client requests data not already on-net—provisioning new capacities will remain a very manual process.

“To go from an inventory of manually deployed services to a more automated environment is a big technical challenge. Everything underlying it needs to be uniform—and in manual processes, they usually aren’t,” Caudell says. “Once network devices start to be controlled by an SDN controller, that tells all the devices what they should be doing… and ultimately, you can get to deploying services in real time because you’re taking away the manual programming level and all the risk associated with it.”

Performance Anxiety

Even harder to achieve—but important to firms active in high-frequency trading or providing direct market access to clients—is the ability to support a firm’s algorithmic trading capabilities with intelligent and automated self-provision and the ability to determine and configure connectivity routes, says Dan Watkins, managing director of consultancy Vantage Point Consulting, who has worked at various network providers in the past. And to do this intelligently, these systems must be able to monitor essential facets of network performance, such as latency, and marry these with the firm’s business strategy.

For example, before deciding whether to adjust bandwidth levels or implement changes to network routes, firms would monitor latency—the delays introduced as data travels from one point to another.

While decisions to re-route trade flow are typically taken by smart order routers—rather than the network level, where sophisticated firms will monitor metrics such as latency and fill rates, and instruct trading algorithms to de-prioritize underperforming routes or markets—these same metrics can be used to evaluate connectivity requirements of new initiatives.

“Where there is more than one carrier, firms want more transparency—much as buy-side firms now expect more transparency from the sell side,” says David Murray, chief business development officer at latency monitoring technology provider Corvil, adding that this burden of monitoring and evaluating performance falls on the client. 

“When you don’t know if a problem is down to your strategy or the network, a systematic trader knows there is a relationship between performance and outcome (i.e. fill rate)—so they need to be able to look at the reliability of the network as well as correlate it to the outcome,” which might influence a network provisioning decision, Murray says. “On a number of occasions, we’ve had customers use us to monitor their exchange connectivity, only to find out that the level of service they were paying for was not what they were actually getting.”

ITRS’ Geneos product monitors network performance, and can issue alerts to trading desks if it spots network components—such as individual switches and routers—underperforming, or even interrupt trading flow to stop trades being sent to market if it detects that a delay would mean that the market could move while the order is still on its way.

In addition, these tools perform a vital function in informing any decision to provision or switch off services. “There has to be a feedback loop on SDNs… so you know whether you have enough bandwidth, and how much you are using,” ITRS’ Warren says. And this feedback loop can be harder to operate than on traditional, static networks. “It’s analogous to bursting to the cloud: as that workflow gets shared across servers and the server workload drops, it means that you have to be watching more servers, each doing less work.”

Though not yet commonplace, Warren says SDNs will inevitably become more widely used. Caudell adds that for capital markets firms cautious about embracing new SDN capabilities, cost, revenue, and client service benefits alone may not sway them—but competitive forces might. “They would say, ‘If our competitors do that and we don’t, they’ll have a big advantage, and we’ll have to do it, too.”