.jpg)
The targets proffered by Matt Simpson, associate director of electronic trading systems architecture at the Chicago Mercantile Exchange, are undoubtedly impressive: a drop in bandwidth requirements of 75% and a reduction in latency from 10 milliseconds to 2 milliseconds, both by the end of 2006.
Mr Simpson hopes to achieve these goals by the implementation of FAST (FIX Adapted for Streaming), the compression algorithm designed to make FIX, the protocol originally designed for trading, feasible as the standard for distributing real-time data.
Of course, not every exchange is as vigorous as CME in its embrace of FIX and its derivatives. Mr Simpson's enthusiasm is partly explained by his role as co-chair of the global technical committee of FIX Protocol Ltd (FPL), the non-profit organisation that seeks to promote FIX as an industry standard.
However, over the last few years, the broader exchange community appears to be steadily moving Mr Simpson's way. FIX, an initiative that is now more than a decade old, is increasingly drawing interest from bourses which have previously largely ignored it.
The way has historically been led by the North American electronic communication networks (ECNs), which, emerging in a highly competitive environment, turned to new technology at the outset to give them an edge. For a broker, FIX lowers the costs and barriers associated with connecting to a new exchange.
For ECNs, the absence of legacy systems made it easier to be in the vanguard of FIX - and they can still be found at the forefront of new developments. Archipelago, now merged with the New York Stock Exchange, was one of the four first adopters of FAST, alongside CME, the International Securities Exchange (the all-electronic US options exchange) and Sweden's OM Exchanges.
But FIX is now steadily percolating down to exchanges who have hitherto shown no desire to sit on the 'bleeding edge<' of technological advance. No fewer than 32 exchanges are now members of FPL - most of these working with, or planning to implement FIX in one form or another.
The explanation for this shift can be found in a wholesale change in market conditions. The inexorable rise of electronic trading in all its forms - including programme and, increasingly, algorithmic trading - coupled with an increase in market participants, many of them hedge funds, has seen exchanges faced with a sharp reduction in average transaction size accompanied by an explosion in trading volumes. For example, the average trade size on SETS, the London Stock Exchange's electronic order book, has fallen by two thirds over the past five years, while annual volume has multiplied more than five-fold.
The increased adoption of quantitative trading strategies has produced growing demand for a greater depth of data, as well as lower latency from data feeds. Further, the latest upward leg in the equity bull market that began three years ago has seen an increase in both listings and the products offered by exchanges.
In short, as observed by a London Stock Exchange official at January's FAST technical summit hosted by the exchange: 'Peak message rates are outpacing communications services and cost-effective bandwidth solutions'. Add the fact that proprietary messaging technologies are increasingly costly to maintain, and it becomes clear that the industry is ripe for a radical shift.
All this is happening in an environment where, though still limited, competition for listings and new trading members is gradually increasing and likely to accelerate.
Like many of its peers, the LSE's stance on FIX is interwoven with a wider technological initiative: specifically, its four-year transition to next-generation IT systems, which last year saw the implementation of Infolect, its real-time information delivery system. The programme, internally titled TRM, or Technology Road Map, is its biggest technology project since the introduction of the SETS order-driven trading system in 1997 and will culminate in the launch of its new version of the trading system in the first half of 2007.
The LSE has already been encouraged by the response to Infolect, which replaced the London Market Information Link (LMIL). The system has three times the capacity of LMIL, currently broadcasts over 10 million messages a day and can be scaled up to 100,000-plus messages per second.
Since it went live in mid-September 2005, the exchange has witnessed 17 of the 20 busiest trading days in its history. One theory is that Infolect's cut in end-to-end latency - from 30-40 milliseconds to averages as low as 2 milliseconds - has stimulated activity by algorithmic trading engines.
The LSE is in now the process of evaluating FAST internally. It plans to offer a prototype FAST market data feed to pilot users from June, followed by performance testing and evaluation through to August before presenting the results in September.
'We are putting our toe in the water', says Wendy Morgan, head of real-time data information services. 'We will seek feedback before moving to a full-blown product offering'.
But the LSE is clearly not alone among the 'second wave' of exchanges exploring the possibilities of FAST. Of the 580 respondents to FPL's recent FIX survey, conducted by Tower Group, 49 were from exchanges. Some 28% of these are currently implementing FAST, while 46% are considering such an implementation. Only 26%have ruled it out.
If and when the LSE goes live with FAST, it expects latency to be reduced even further by the efficient way the algorithm compresses data. The exchange concedes its primary audience is those customers whose internal systems are already geared to FIX. 'If we give them the option of using FIX natively in connecting to us, they will use it as they can eliminate a layer of translation', says Morgan.
The launch of TRM should also offer clients the option of connecting to the LSE either through its existing, proprietary SETS interface or through FIX. To accommodate potential differences in latency, the exchange plans to conduct extensive tests, publish the results and advise clients on the best solution for them. 'We don't expect the FIX connection to be any slower but, even if it is, those customers whose internal infrastructure is based on FIX might reduce latency by avoiding the need to translate between protocols. Taking out one layer of protocols at the client's end will also have a positive impact on staff and software costs'.
But the one explicit benefit that most exchange users of FIX are hoping for is the potential of the FIX protocol to help in attracting new members. For example, the LSE says it has firms waiting to join who want to use FIX to route orders from the US, but who must in the interim either wait for TRM to go live, or use the proprietary SETS interface.
Similarly, the Warsaw Stock Exchange cites the lure of increased membership as one of the strategic reasons behind its implementation of FIX, which provides trading access to its central systems. The exchange went live with FIX in March 2005 after nine months of preparation.
Jurek Kleszcz, who acted as senior adviser on the WSE project, describes the FIX implementation as, 'the provision of a more attractive, lower cost, interface that will help the exchange's drive to acquire new members'. Flexibility, or, as he puts it, 'future choice for technological changes for both the exchange and its members', was also critical. In addition, 'embracing FIX was seen to help Warsaw's image as the most technologically advanced exchange in the region'.
Another reason for implementing FIX is as a springboard for new services. The London Stock Exchange is seeing interest from buy-side firms wanting to use FIX as its mechanism for connectivity. 'The buy-side has been very supportive of our efforts to implement FIX', says Morgan. When trading, buy-side firms would have to be 'sponsored by member firms offering a DMA (Direct Market Access) style service', continued Morgan. 'The service's aim is to increase speed of access, not to disenfranchise brokers'.
Elsewhere, virt-x initially employed FIX as a means to lower the costs of connectivity for new members. It first went live in 2003 with FIX, which it provides alongside its proprietary interface as an alternative aimed specifically at low-volume users. Two vendors - Bloomberg and GL-Trade - consolidate order flow from their clients and send this to virt-x using FIX. In addition, 'about a dozen' members, out of 110, connect using FIX. virt-x now regards the protocol's evolution and its adoption among exchanges as an issue of strategic importance. 'We keep an eye on developments, particularly FAST', says Patricia Lynch, Director of Market Relations and Sales.
One facet of FAST, alluded to by CME's Simpson, that is likely to appeal to most exchanges is its efficiency: particularly its lower bandwidth and processing requirements. A pilot conducted by OM Exchanges using its SAXESS electronic trading system found that the use of FAST achieved 65 per cent better bandwidth utilisation and made more efficient use of central CPU resources by a factor of between five and 10 times.
Not all exchanges are looking at FAST, however. Euronext has used FIX since 2001, when the pan-European bourse was created from the merger of the Amsterdam, Brussels and Paris exchanges. It implemented the protocol at the same time as launching its NSC cash trading system, and has offered it alongside its proprietary Market Message Transfer Protocol (MMTP), which was developed six years ago.
Anthony Attia, head of operations, cash market and listing, at Euronext, says, 'We do not have any dogma in terms of message protocol. We are entirely pragmatic in following the requirements of our users'.
'The strong demand from our members is for stability and performance. All changes require investment, and this is something they wish to avoid unless a good case can be made for it'. Rather, says Mr Attia, Euronext has focused its efforts on a series of improvements to its central trading systems - both software and hardware upgrades - that have cut latency by 40 per cent over the last 18 months.
Deutsche Börse is also somewhat equivocal when it comes to FIX. Like most of its counterparts, it has pursued a twin-track approach, offering members use of either its proprietary Values API or its Xentric FIX Gateway. The latter, unveiled two years ago, allows members to connect FIX-based applications to its Xetra and Eurex trading systems.
However, a Deutsche Börse spokesman expresses reservations that FIX has grown up as a protocol between the buy and the sell side, and has not been explicitly tailored for the needs of exchanges. Further, he cites concerns among members at the divergent versions of FIX currently in use. 'There is room for clearer harmonisation as FIX moves forward', he says. 'In this regard, we are working quite closely with other exchanges'.
If FIX it is to maintain its recent impetus among exchanges, it may find further momentum from the ever-present spectre of cross-border consolidation. In this context, the use of FIX may play a role in smoothing the path of future linkages between exchanges, whether through outright merger or otherwise.
Most importantly, it would enable member firms on both sides of a merger to avoid the requirement of writing to a new proprietary interface. 'Both exchanges could standardise on FIX to reach the client base, effectively hiding the work that has to happen in the background to merge two systems', says an exchange official who did not wish to be identified.
