Blog article
See all stories »

Ultra-low latency clouds - horses for courses

What happens if “The Cloud” doesn’t deliver what you need?  One answer is to find a cloud that does deliver what you want.  With the world being given the impression that “Cloud” means “the public Internet”, many folks aren’t aware of some of the non-Internet cloud approaches that major investment firms around the world are taking.

What is a “cloud”?  Let’s call it “a network platform shared by multiple service providers and multiple service users with a single connection from each participant location”.    Like any cloud, the edges of that definition can be a bit fuzzy – some users would be using so much capacity that they need or want more than one connection to that cloud, or they might want multiple connections for BCP policy reasons.  But the concept of a “single connection” to a cloud is a key differentiator.

The public Internet doesn’t meet all of the reliability and security requirements of financial institutions, which is why “private managed clouds” have been built for the financial community – they are “managed” to meet reliability requirements and “private” to meet security requirements.  The public Internet and private managed clouds have typically used the Internet Protocol as a standard, which can allow services provided by banks, brokers, exchanges, etc to be virtually “plug-and-play” compatible for delivery over both types of cloud.

But the largest firms are now starting to use private managed clouds that work at a layer below the Internet Protocol in order to meet their ultra-low latency requirements, and particularly for connecting between the hosting centres in or near major financial centres that are used by exchanges, ECNs, MTFs and trading platforms for hosting their trading systems.

So, next time that you hear the term “Cloud”, don’t think that that just means “the Internet”.


Comments: (1)

A Finextra member
A Finextra member 18 January, 2013, 12:20Be the first to give this comment the thumbs up 0 likes Chris, One of the differences between a "cloud" and a "central server" is ubiquity of connectivity. In most people's minds, I guess, that relates to GSM or WiFi. HTTP is the most common protocol in either case. There could be several causes for latency, but if you have just 2G connection in a cell (i.e. not 2.5G aka GPRS), guess what - you have only one option for data connection (which happens to be relatively low-latency too). As for the financial transactions and latency, the bottleneck is often not at the data comm channel level.