1. Network Ultra-Low Latency: how consistent?
A classic sales strategy in today's network market is to headline some fantastically low latency number. But does that number actually refer to the provider's entire network? Or just a carefully cherry picked segment between carrier Points of Presence (PoPs)? Unless all your traffic will only ever run on that segment, the latency number can be incredibly misleading. Make sure you read the fine print and know your end-to-end points.
And perhaps more importantly, what happens to latency performance in a failover situation? If the provider's back-up line exhibits latency ten times that of the primary, then your trading operations may be as badly affected as if you had no connectivity at all.
Lastly, have a point of reference. It all comes down to speed and
accuracy. As an example, industry leading latency numbers are sub
66 ms RTD from
New York to London and 10 ms from Chicago to Toronto. Highly reputable carriers, like Hibernia Atlantic, will ensure your latency performance with Service Level Agreements (SLAs).
Joe Hilt & Fergus Innes
2. Time to market: how long will my install take?
Obviously network latency is a big concern for any trading
operation these days but the time for circuit turn-up of a
provider also has a huge effect on your profitability. A classic
example is installation time; a carrier may be touting sub
microsecond latency, but why is it going to take his firm 30 to
90 days to install the circuit? As a guide, Hibernia guarantees 5
days from signature to service. Make sure you ask the turnaround
time before you ink the deal.
Remember, the provider install timeframes directly affect your P & L. You may have spotted a great trading opportunity that requires some extra connectivity but every day you're waiting for fibre is another day that someone else can beat you to it.
3. Network capacity: when is my bandwidth not my bandwidth?
So a provider is offering you 'X Gbps' Ethernet, but is that completely non-contended? It's alarmingly common for providers to sub-lease a 10 Gbps circuit, resell it as 40 x 1 Gbps circuits and hope that their customers won't want their maximum bandwidth simultaneously. Any provider with a clue about financial markets knows that this math doesn't add up; when markets kick off, data rates go through the roof and all financial customers will need all their paid capacity. So make sure that the 'X Gbps' Ethernet you're being offered is truly a dedicated private network for your company and not shared.
4. Network flexibility: I need more, when can I have it?
Another litmus test of a network provider's financial market expertise is rapid flexibility. The continuing growth in financial message volumes and transactions nearly guarantees that today's 100 Mbps circuit will need to be tomorrow's 1 Gbps. That situation prompts two important related questions - one local and one pan-network.
At the local level, a financial networking provider will always try to future-proof an initial install. They'll commission the circuit with potential capacity over and above the initial requirement so that any future upgrades are fast and non-disruptive to your business. Providers who don't do this will cost you money and aggravation later.
At the pan-network level, there's the question of overall
capacity. If your trading operation needs extra bandwidth to cope
with higher message volumes, it's likely that your peers will
too. In which case, how will your provider handle that
at the macro level? What wavelength capacity are they running on their core routes, such as across the Atlantic? Are they still using just 10 Gbps wavelength capacity or have they already made the move to 40 Gbps? Is the provider that you're dealing with innovative and technology driven? If they are still on 10 Gbps, what percentage utilisation are they already at on that core route? 35% or 95%? Always make sure there is a solid SLA (Service Level Agreement) in place that protects you.
5. Cable: whose is it?
A critical question that affects many aspects of a provider's performance is whose cable is delivering the service? Many providers are resellers or sub-resellers of somebody else's fibre; they have no direct physical control or view over any part of the network on which they are selling service.
That obviously raises concerns around service levels; an outage
that affects your mid-sized or small reseller may not be the
cable owner's top priority. There's also the question of
expertise; if you're dealing direct with the cable owner, you can
be certain you're also dealing with the sort of people who eat,
breathe and live the networking business - not someone in a call
Additionally, if the bandwidth is sub-leased, technical service can become infuriating. If you are dealing with the direct cable owner, who has diverse and global Network Operating Centers (NOCs), responsibility is clear and troubleshooting is fast. Another big question you have to ask is did your reseller recently pay the leased capacity invoice to the cable owner for your service?
By the same token, cable ownership also affects pricing; if you are dealing with a provider several middlemen distant from the cable owner, the price you pay has to cover multiple layers of costs and profit margins.
Historically, these disadvantages have been a fact of life for some financial customers, because their requirements were too small-scale to interest the owners of major networks. That is no longer true; it is now possible to lease capacity direct from a major wholesale provider/owner and benefit from the same pricing and service levels as the very largest media companies and financial institutions.
6. Cable: where does it go?
The natural tendency in many financial firms is to think of redundancy in the most local sense; e.g. are the cable runs from different providers into our building sufficiently segregated in terms of entry points and ducts? However, that overlooks an equally important bigger question, namely what segregation measures do network providers have in place elsewhere? If the carrier you choose does not have physical diversity throughout its network footprint, then the carrier is putting your critical information at undue risk.
Some common industry practices are decidedly alarming. Sharing common backhaul routes may save carriers money, but also increases operational risks for their customers. One classic example of this is on the southwest coast of the UK where four of the largest transatlantic submarine cables all run together for a considerable distance in the same eighteen inch wide duct! A similar situation applies in North America in both NJ and Long Island, New York. All three major cable stations in NJ use the same common backhaul up the Garden State Parkway (GSP) into the New York Metro area. In Long Island, the same congested backhaul event occurs but on aerial telephone poles into New York City. .
Routing through major centres is another potential issue. Some providers offer capacity on networks that have a point-to-point architecture - so an outage in a city will interrupt all traffic flowing through it, not just local traffic. More professional providers will have built bypass circuits around major cities to avoid this, in case of local or global emergencies.
7. Conclusion: is the carrier financially savvy in its transport services?
Many of the questions above derive from one 'master question': Is transport for financial firms your core focus? A firm that has financial market connectivity at its core will be thinking differently from the average reseller. Rather than just reacting to industry changes, it will anticipate them by analysis. Which major trading platforms are planning to open new data centres? Do you have diverse connectivity to local and overseas exchanges? Where? What's changing in the trading demographic? How? The net result is that any financial firm can be confident that an enterprise-savvy provider will always be ready with the right capacity, resilience and latency, in all the right places, at the right time. Hibernia Atlantic and the Global Financial Network (GFN) seem to tick all the boxes.
For more information on ultra-low latency connectivity, email firstname.lastname@example.org.