Low Latency Fiber Routes To Europe and Asia

Many US companies have offices in the UK and Europe. Multinational companies have their operations located worldwide. So, how do you reliably interconnect all of your international business locations including connections to the cloud?

More and more, the answer is low latency fiber routes. Low latency is a term you hardly ever use to hear. It became part of the user vernacular with the switch from circuit switched phone service to VoIP. Before that, latency was a derogatory word used to describe the big delay experienced when communicating through geosynchronous satellite links. Now all the major carriers are touting their new low latency fiber routes to Europe or Asia.

So why the big change? Aren’t fiber routes inherently low latency?

Indeed, most fiber optic connections have latencies in the tens of milliseconds or less. That’s enough to ignore for most purposes. Notable exceptions are high speed financial trading, data replication and some cloud services.

Don’t confuse bandwidth and latency. Bandwidth, often referred to a line speed, is how fast you are sending out ones and zeros. This may be limited to the capability of the terminal equipment at either end or, more often, the capacity of the line. You can connect Gigabit Ethernet routers to each end of a T1 line but your packets will transfer no faster than 1.5 Mbps. That’s the T1 line speed.

Latency is a different animal. It describes the time it takes for those ones and zeros to get from one end of the line to the other. Often latency is quoted as the time in milliseconds for a round trip from source to destination and back to source.

Here’s how latency rears its ugly head. Say you are trying to carry on a conversation with someone using VoIP on a connection that has several hundred milliseconds of latency. That sounds like a lot, but it’s not uncommon on the Internet. The two way conversation quickly becomes like using a walkie-talkie. You talk one at a time and then pause for second before you speak when the other person is done. If not, you find yourself talking at the same time, but stepping on each other’s conversations.

More to the point of fiber optic latency, say you and someone else are trading stocks using automated programs. If your algorithms both decide there is a buy opportunity at the exact same moment, they’ll each place a buy order of a certain size based on the price action they are detecting. What happens if your order reaches the exchange 10 or 15 milliseconds after your competitor? It’s entirely possible that they’ll get a better price because the market moved between the trades.

This is what is getting “quant” traders exercised over latency and why there is a substantial market for the lowest latency connections possible. Carriers can reduce the latency of their fiber optic routes by taking the shortest path possible between two locations and minimizing the amount of equipment in that path. Every switch, router and multiplexer takes a small amount of time to do its job. Add them all up and you’ve increased latency.

Ultimately, the speed of light through the glass fiber is the limiting factor. That’s why shorter paths are lower latency. For the ultimate in latency reduction, you want to colocate in the same facility that houses the exchange computers or very near by. In high speed trading, even microseconds now count.

Latency was never that much of a problem when the bulk of traffic was one-way data file transfers. They get there when they get there and TCP/IP makes sure they arrive intact. It’s also no big deal for one-way audio and video streams or downloads. You just have a slight delay before the program starts. You probably won’t notice latency because there is a much larger delay built-in to buffer for jitter caused by congestion and packets taking different paths through the network. Two-way voice (VoIP) and video (teleconferencing and telepresence) are definitely affected by latency, although not to the same extent as financial trading and other very latency sensitive processes.

Does your company need higher performance network links to be more effective? You’ll do well to review the cost/benefit tradeoffs of low latency fiber optic networks, MPLS networks and other options available for your business locations.



Note: World map image courtesy of NOAA on Wikimedia Commons.

GTT Expands Ethernet Footprint With PacketExchange

Global telecommunications carrier GTT is moving to expand its Ethernet service footprint worldwide by acquiring Ethernet service provider PacketExchange.

PacketExchange is a London based carrier with an impressive 10 Gigabit fiber optic IP backbone network that covers Europe, the United States and Asia. A good portion of this network was acquired when PacketExchange acquired MZIMA Networks in 2010. The combined resources provide private line services plus MPLS and VPLS networking and peering.

GTT’s acquisition of PacketExchange reinforces the idea that we’re entering the age of Carrier Ethernet. Up till now, the world’s networks have been dominated by telephone company technology for switched circuit operations. The popular T-carrier wireline services and SONET/SDH fiber optic services were all devised to be compatible with time division multiplexed voice and data. The new paradigm is packet switched networks, with Carrier Ethernet gaining prominence.

You see this reinforced by the emergence of Ethernet Exchanges, such as Telx, that offer E-NNI (Ethernet Network to Network Interface). What E-NNI does is create a peering opportunity for IP network operators to exchange data and, thus, expand the reach of their networks. This opportunity speeds up the ability of Ethernet to take over the world’s telecommunications traffic without the necessity of one dominant international carrier to dictate standards.

PacketExchange makes a nice complement to GTT which is already a global network integrator with many POPs (Points of Presence) in the United States, Europe, Asia, Australia, Africa and South America. GTT serves 80 countries in all with a wide portfolio of Private line, Ethernet, DSL broadband, point to point Wireless and managed network services. They are often the carrier of choice for companies that need to interconnect offices across international borders. MPLS network services that are particularly useful in linking multiple locations include MPLS VPN and VPLS services.

Ethernet and MPLS networks complement each other nicely. MPLS networks use a proprietary routing technology called label switching in place of IP routing or SONET/SDH multiplexing. Label switching encapsulates virtually any protocol and transports the packets form location to location. At the network edges, packets enter and leave the network unaltered. This allows MPLS networks to support layer 2 switched services such as Ethernet Private Line and Ethernet LAN. It’s entirely possible using Ethernet access to MPLS networks to create an international bridged LAN that includes locations worldwide.

The power of Ethernet as a universal network protocol to support converged voice, video and data networks is only beginning to be seen. It’s a natural development, considering that corporate LANs long ago standardized on the switched Ethernet protocol. What makes more sense than to seamlessly expand those LANs into the metropolitan and long haul connections necessary to support regional, national and international business.

Does your company have a need to connect multiple business locations, domestically or internationally? If so, take a closer look at Ethernet business bandwidth solutions to support your organization’s technical goals at a cost often lower than with traditional telecom services.



Note: Ethernet switch photo courtesy of Justin Smith on Wikimedia Commons.