What has historically been the exchange's technology strategy as regards its trading platform?
Until 2001 we were using a platform called ASTS that we ran in-house ourselves, but we then revised our technology philosophy to one of sharing a trading platform with strategic partners to maintain a wider network for distributing our market. To that end, we went live with SAXESS from NASDAQ OMX in 2001 and ran that until 2010. In March 2009 Oslo Børs and LSEG entered into a strategic partnership to co-operate across their equity, fixed income and derivative markets. As a part of this agreement, we switched in April 2010 to the LSEG's TradElect platform. The main reason for changing to TradElect was that we saw more potential in a strategic partnership with the LSEG in that we would be able to attract more members by being on the same technical platform as London.
What was the thinking behind the latest announcement of switching to Millennium Exchange so soon after moving to TradElect?
The LSEG acquired MillenniumIT very shortly after we went live with TradElect and deployed Millennium Exchange on Turquoise about a year ago. Since then, the main market in London has also successfully migrated to Millennium Exchange, while Milan and Johannesburg are now in the process of doing so. It didn't make sense for us to remain on the previous technology in view of this migration trend - and also in view of our desire to make it as easy as possible for members to connect to us.
From a purely technology perspective, what are the aspects of the Millennium Exchange platform that appeal to you?
Resilience and performance. From purely a technology perspective it also appears a really well designed and modern platform. It is easy to configure and it is possible to apply more logic at a configuration level than our current platform. It also has a far more compact technology footprint than TradElect and it is possible to run it on a smaller stack of servers, so it is a leaner and more agile platform. It also runs faster for a given hardware configuration and it appears to be relatively straightforward to scale the platform quickly and add capacity, which is another attraction.
Latency has obviously become an ever more important factor for market participants, and the implementation of Millennium Exchange will strengthen our competitive position in this respect. The system is also flexible, which will make it easier for Oslo Børs to implement changes to respond to market developments.
There is also the issue of cost. As a relatively small exchange, Oslo Børs needs access to market leading technology, but at an affordable price. The only practical way for us to achieve this is by teaming up with larger entities, such as the LSEG. It simply wouldn't be feasible for us to acquire low latency infrastructure of this quality any other way.
Where will you host your matching engines when you go live with Millennium Exchange?
Our matching engines are currently hosted in London and this will probably remain the case. However, any final decision on this has to accommodate the wishes of our members. Our local members would obviously like it to be hosted in Oslo, but our overseas members would prefer to stay in London. Clearly, that is a business rather than a technology decision, but from a technology perspective we could deliver either option.
What instruments will you be using Millennium Exchange for?
We will be using it for the same instruments currently traded on TradElect: equities, fixed income, warrants, ETFs and ETPs. All our derivatives markets will remain on the SOLA platform, which we also have in common with the LSEG/Turquoise.
Will the use of the new platform make it easier or faster for the exchange to launch new products?
It will probably make it easier from a market supervision perspective because the client tools are high quality and easier to use than those on TradElect. It is also probably easier to use in most respects than TradElect.
What testing are you undertaking as regards integrating the new platform with your in-house technology systems?
We have a trial version of Millennium Exchange that has been running here on a couple of servers in Oslo for a while. (Incidentally that sort of trial deployment probably wouldn't have been practical in the case of TradElect because of its much larger hardware footprint.)
This trial version of the platform gives our internal developers a development environment with which to become acquainted with the system and to facilitate their building of trading system adapters.
What hardware and operating system combination will you be running in production?
We'll be using commodity Intel blade servers running SUSE LINUX (which is also the configuration we currently have on the test systems in Oslo) but the platform could also be run on Solaris or other operating systems if required.
We're pretty agnostic and pragmatic about operating systems, with a focus on rationalisation and minimising potential maintenance issues in view of our compact IT team. Until a couple of years ago, we were running a mixture of operating system technology - UNIX servers, LINUX, and some Windows. However, all our internal systems at the exchange (what we refer to as our downstream systems) are now Java based and run on Windows, largely for cost reasons. Our current TradElect platform is based on .NET and we communicate with that through gateways and adapters, so it is pretty much transparent for us.
You mentioned "trading system adapters" earlier. Can you elaborate on these?
We will be developing Java-based adapters to give us a private data feed from the trading engine, which we will be using for internal activities such as index production and our data warehouse. We will also develop adapters for scrutinising market data feed channels such as FIX and there will be a further adapter via which we can send news messages and calculated index values back into the trading system. We already have similar adapters in use today with the TradElect platform for calculating and delivering index values, which we do in Oslo rather than in London because it is a less latency sensitive task.
Millennium Exchange comes with a number of external gateways, so it is fairly easy to develop this sort of adapter. The platform also provides functionality and gateways that we currently don't have in the TradElect system, such as drop copy and CCP gateways. So it is more extensible in that respect.
Are indices a significant part of the exchange's business model?
Yes. Our index engine in Oslo is used for a lot of other products based upon index calculations, such as custom indices. These custom indices are part of our product portfolio in the market data area and we provide them to third-party clients (such as our members) who wish to have a certain index calculated by the exchange to their specifications. This has long been an important part of Oslo Børs's business and is an area where we have a well-established track record.
Again, all this is undertaken using Java running on Windows. All the in-house databases and feed handlers that support these index calculations (and other internal activities) are fed by a real-time Tibco Rendezvous message bus.
Is all your Java systems programming undertaken in-house?
Yes. The in-house development team has been pretty stable now for a couple of years and it already has plenty of experience in writing adapters for the TradElect system, so we have a fairly strong base of experience in that area. The programmers working in the development team are all exchange employees, not contractors.
Have you noticed a change in the exchange's participant demographic in recent years in terms of either location or trading style?
There has definitely been a change in our membership, in that approximately half our members are now based outside Norway. We have also noticed that we have been attracting appreciably increased high-frequency trading activity since we went live with TradElect in April 2010.
Do you find that your dialogue with members on technology matters varies depending upon their location?
No. The topics tend to be more or less the same - latency, co-location and smart order routing. These three topics tend to top the list, though the last has become more prominent recently as an increasing number of our members have started to use some form of smart order router.
Has the exchange's changing participant demographic been reflected in the nature of its liquidity?
Absolutely. We have definitely seen a dramatic change in the order to trade ratio since April 2010, with far higher order message volumes than in the past, with average trade sizes also becoming smaller. However, at present it is difficult to say decisively whether or not actual trade volumes have gone up or down, because recent volatility in the market due to factors such as the Eurozone debt crisis have distorted the picture.
What about message to trade ratios? Do you charge if participants exceed certain thresholds?
We are aware that some other exchanges do this but at present we do not. However, we are always considering and discussing different pricing models in order to attract as much liquidity as possible.
Increasing message traffic volumes obviously have a cost implication for the exchange in terms of network bandwidth. We maintain a dedicated backbone network between Oslo and our matching engines in London, so if order volumes rise sharply we would have to expand bandwidth, which costs money. However, from a technical perspective it is mostly UDP multicast traffic so it is relatively easy to manage and monitor on a day to day basis.
Does the exchange offer co-location facilities?
Yes. We offer our markets through the same co-location partners (such as Colt) that the LSEG uses. Having our matching engines and co-location facilities based in London makes us more accessible for those firms trading across multiple markets who have the convenience of being able to do so from within one facility, or at least across multiple facilities that are in close proximity to each other.
Finally, what other technology projects does the exchange have in the pipeline at present?
As a relatively small exchange we tend to focus on one major project at a time, so the migration to Millennium Exchange will be our primary focus once it gets underway (probably later this year). However, in the meantime we're running a project to simplify and scale down our internal systems as much as possible. More specifically, a major task at the moment is downscaling and modernising our database and data warehouse systems.
We are currently rewriting the data warehouse system in line with our philosophy of having as small a technology stack as possible. We do not like big enterprise tools that we cannot utilise fully, so we try to base as much as possible of our internal system portfolio on the same technology. We have a limited resource pool, so we have to make sure that our developers are able to support the main elements of the system portfolio that we run.
The new version of the data warehouse will be based upon SQL Server, with Business Objects being used as a reporting tool. We will be reducing the number of ETL tools used, while increasing the use of Java programs for aggregating values. Everything will be run on Windows using commodity hardware (we have standardised on HP blade servers). As I mentioned earlier, we previously also ran a variety of other hardware and operating system combinations, but when you have an in-house technology team of barely 30 people you cannot afford the luxury of multiple platforms!