Presentation on theme: "1 WFE MIT Forum Comparisons of ATS systems with those of traditional Stock Exchanges Trading Technology November 2009."— Presentation transcript:
1 WFE MIT Forum Comparisons of ATS systems with those of traditional Stock Exchanges Trading Technology November 2009
2 AGENDA The new exchange participants Architectural Differences ATS system architecture Challenges to the industry
4 Industry Trends – exchange participants Global brokers Simple connections, FIX Routing networks Co-location Cheapness Product expansion Commodities, Derivative, ETFs. Local Brokers No changes Cost reduction Simplicity Algo Traders: Cheap Fast High volume/low latency Co-location Simple market structures Out of hours trading Not afraid of technology!
5 Industry Trends – effect on execution centres In 2003 trend was BCR centric – resilience and recovery key High degree of functionality – derivatives, ETFs etc Multi-asset/functionality systems Scalability – Moores law kept up with increase in business In 2009 trend is Low latency Performance High order to trade ratio Competitive costs – low overheads
Industry Trends – Architectural Types 7 TT has identified four major types of architecture for exchange trading systems. Type 1.Mainframe/minicomputer centralised trading system 2.Distributed multi-server resilient system 3.Simple (simplex) trading system with few or no resiliency components 4.Web/windows component systems These are generally in order of chronological development There is no judgment as to which is better or worse The packages such as CLICK, X-STREAM and SAXESS are type 2 The ATSs tend to be a type 3, which has emerged as the highest performance system due to its simplicity
8 8 Typical Type 2 system architecture
9 9 Typical ATS system architecture
10 Speed = simplicity...
11 Industry Trends – Trading systems ATS model – simple systems Designed for speed first, resiliency second Sacrifice resiliency components for speed Reliant on more stable platforms HP x86 Blade servers Clustering Disk RAID/SAN resiliency Use of Multicast core to architecture Use of native protocols plus ITCH OUCH Standardised on C++, RedHat Linux, Open source components, MYSQL or Oracle Offer Co-location
Trading system latency Note these are published latency figures Beware lies, damned lies and Exchange statistics! Beware also that all technology providers will claim sub nano-second latency... ASX is spending money on upgrading ITS BATS and Chi-X will not be standing still on latency Will Latency matter once all venues are under 1ms?
15 Industry Trends – effect on execution centres Added together, 86% of business in the top 25 exchanges goes through an ECN-style system
16 Industry Trends – Effect on execution centres In a theoretical world, and with routing, with all exchanges offerings being equal, market share will tend towards 25% (if there are four participants) Starbucks comparison Thus the market will become saturated...
Industry Trends – Business Strategy Saturated market competition mode is thus Minimal overheads Maximum automation Maximum business throughput Possible cross-subsidy of main business streams (e.g. listings and market data to subsidise trading) IT a core role Needs to be minimal 17
18 6. Conclusions
19 Summary – future state architecture? Has the time between trading system redevelopment shrunk? Used to be 10 years Designed for speed first, resiliency second Fewer resiliency components – be brave! Use of Multicast core to architecture Use of native protocols plus ITCH OUCH Standardised on C++, RedHat Linux, Open source components, MYSQL or Oracle Offer Co-location