No changes Cost reduction Simplicity
Cheap Fast High volume/low latency Co-location Simple market structures Out of hours trading Not afraid of technology!
• Algo Traders:
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 – Moore’s law kept up with increase in business
• In 2009 trend is
• • • • Low latency Performance High order to trade ratio Competitive costs – low overheads
5
1. Architectural differences
6
6
Industry Trends – Architectural Types
• 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
7
Typical “Type 2” system architecture
8
8
Typical ATS system architecture
9
9
Speed = simplicity...
10
10
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
11
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?
Equity Trading System Latency (ms) - 2009
25.0 20.0 15.0 10.0 5.0 0.0
) ) ) ) ) y) c t) ss tra XT ary ar e t e tum e t l r k e n e e p ri (X ri lic ua ad Ex G op (C rop r ( tr (Q B e p p X ( d ( D E X ra AS i-X LS TS TM (T h A a C B ph Al
6. Effect on the industry?
13
13
13
List of top 25 Exchanges*/ATSs using ATS systems
Rank
1 2 3 4 5 6 7
Exchange/ATS
NASDAQ OMX (US) NYSE Euronext (US) BATS (US) London Stock Exchange Tokyo Stock Exchange Deutsche Borse NYSE Euronext (FR)
Toronto Stock Exchange
HK Exchanges Borsa Italiana Swiss Stock Exchange Korean Stock Exchange Stockholmbörsen Australian Stock Exchange Shenzhen Stock Exchange Taiwan Stock Exchange BM&FBOVESPA National Stock Exchange of India American Stock Exchange Tadawul (Saudi Arabia) Oslo Børs JSE Securities Exchange
Industry Trends – effect on execution centres
• Added together, 86% of business in the top 25 exchanges goes through an ECN-style system
Traiditional SE systems ECN system
15
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...
16
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
6. Conclusions
18
18
18
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