Tuesday, October 17, 2017

‘Making split second decisions’: How high speed optical networks are revolutionising financial infrastructure

Data Economy speaks to Koby Reshef, CEO of PacketLight Networks, on how high speed optical networks are changing mission critical analysis and transactions in financial markets.

The financial sector is one where speed is important as the minimal latency could case damages worth billions of dollars to some of the large enterprises and stock exchanges.

High speed optical networks are being laid out at an unprecedented speed to answer the rising demand the industry is seeing from all corners of the world, from Sidney to Sill icon Valley.

Speaking to Data Economy (DE), Koby Reshef, CEO of PacketLight Networks, explains what is happening and what to expect from the network space in the coming months.


DE: How are high speed optical networks changing mission critical analysis and transactions in financial markets?  

The nature of finance trading is about making split second decisions,” Koby Reshef, CEO of PacketLight Networks

KR: The nature of finance trading is about making split second decisions with nano-second accuracy, and it is imperative that the networks that transfer information across long distances gets to users immediately or at least quicker than a competitors network. But it is less about high speeds, and more about achieving low latency.

Financial users buy high speed ports only because 10G ports have lower latency than a 1G port. There are a couple of factors that set the latency of the link.

First, the fiber itself and speed of light inside the fiber, which is 4.7us/Km, is a physical factor that can’t be changed. Because of this, there is a constant look for shorter fiber paths (routes) to reduce this latency.

The second factor is optical transport equipment latency. For long distance optical transmission, it is mandatory to use FEC (Forward Error Correction) algorithm to enable errorless link performance. The FEC introduces latency due to the buffer it uses for fixing errors that occur over the link/distance.

The better the algorithm code is, the lower the latency of the link. This is a factor that equipment vendors can improve and differentiate.


DE: What sort of latencies are we talking about in the financial sector? What is acceptable and what is not? 

KR: Every micro second counts. For short distances, the WDM equipment should introduce sub us (micro second) latency end to end (equivalent for 150 meters). For long distances, since FEC is mandatory, the latency is about 10us (2Km of fiber, which is very small compared to 1000Km of link).

The transactions are automatically performed by machines, so if your network latency is lower your transaction will be registered before the competitor at better rates. There is a lot of money that you can earn selling or buying at better rates.

For example, in New York City most trading houses are located as close as possible to the stock exchange because every meter counts and the selection of the fiber provider is made based on the actual fiber path it can provide.


DE: How fast can we go today and how fast will it be possible to go in the future? 

KR: It is latency dependent rather than the speed of the optical interface. In the future, better FEC algorithms that cope with higher error rate without adding latency will enable longer distance for optical transport and reduce the number of regenerators on the path, enabling faster transmission.

There is also research around different fiber material that will bring the speed of light closer to the speed of light in vacuum, which is the maximum possible.


DE: How do you build this financial network infrastructure? 

KR: Building the infrastructure begins with finding the best (shortest typically) fiber routs available, coupled with optical amplifications sites across the path to avoid the need for regenerators which add both latency and cost.

Then based on the capacity needed, whether it is 10G or 100G, you need the right optical transport solution that aggregates the traffic over the fewest wavelengths as possible.

For long haul networks 100G is the most common capacity of a wavelength, and with the coherent technology, single spans of 2500KM are achievable without regenerators.

This is compared to 10G direct detection technology which is limited to 1000Km or less and requires the use  of dispersion compensator modules that add latency.


DE: What sort of advancements can we expect in the fiber space in the coming years?

KR: There are emerging sets of regulations and practices due to growing threats that financial segment must comply with. The main trend we see is around security and encrypting the data that passes between the data centres.

Since encryption is required for the link, it introduces an additional latency factor. There are few solutions for encryption such as IPSec in layer-3, MAC Sec in layer-2 and Layer-1 optical encryption.

Layer-1 optical encryption is the only solution in terms of latency that is good enough for the trading applications, as it introduces the lowest latency to the link and does not reduce the link speed of capacity.

We expect there will continue to be improvements to security at the different layers that will find a better balance between latency and security.