6. Access Procedures
Financial Controls to prevent:
o Orders that exceed appropriate pre-set credit or capital thresholds
o Orders that exceed appropriate price or size parameters
Regulatory Controls:
o Prevent entry of orders unless there has been compliance with all
regulatory requirements
o Restrict a person’s access to only those market-access systems
and technologies that are pre-approved by the brokers
o Provide surveillance system of the brokers with immediate post-
trade execution reports
8. Implementation options:
• Hardware implementation:
Firewalls
Microcontrollers
FPGA vs. ASIC
• Software implementation:
A program integrated into OS
9. FIREWALLS
Types of traffic:
Packets
Connections
State
Technologies in Firewalls:
Packet filtering
Proxy firewall
Stateful packet filtering (Dynamic Filtering)
Dynamic Filtering:
The best combination of performance and security
ACL (Access Control List)
A state table
10. Microcontroller
• central processing unit - ranging from small and simple 4-bit processors
to complex 32- or 64-bit processors
• volatile memory (RAM) for data storage
• ROM, EPROM, EEPROM or Flash memory for program and operating
parameter storage
• discrete input and output bits, allowing control or detection of the logic
state of an individual package pin
• serial input/output such as serial ports (UARTs)
• other serial communications interfaces like I²C, Serial Peripheral
Interface and Controller Area Network for system interconnect
• peripherals such as timers, event counters, PWM generators, and
watchdog
• A clock generator - often an oscillator for a quartz timing crystal, a
resonator or RC circuit
• A lot of microcontrollers include analog-to-digital converters, some
include digital-to-analog converters
• in-circuit programming and debugging support
11. FPGA vs. ASIC
(FPGA) Field-programmable gate array
Configurable by customer.
Reprogrammed as many times as customer needed.
High performance.
High costs.
(ASIC) Application-specific integrated circuit
o Configurable by customer.
o Non possible to reprogram.
o The performance higher than FPGA.
o The cost lower than FPGA with large amounts.
12. Non-physical implementation. Only a program
integrated into Client workstation
Low costs
Simple installation
Acceptable speed performance for most tasks
Mobility and speed when requirements change
13. Pricing module
Goal: to provide correct parameters to the risk
control
Risk exposure of portfolio
o e.g. options delta calculation
Pricing threshold
o Parameters for pricing threshold calculation
To make it fast:
o Delegate complicated calculations to GPU
14. Concurrent Computing in GPU
Portfolio: Parallel pricing for several options
Monte-Carlo simulation: Parallel simulation of several paths
Delta: Parallel pricing with different market rates
Pricing
Ma
rk e
tR
at e
s
Portfolio revaluation
New Order
Risk exposure
o
tfoli
Por
15. Example: SABR model for American options
(Monte-Carlo simulation)
SABR – Stochastic Alpha, Beta, Rho
o Stochastic volatility model with three parameters – alpha, beta, rho
o Price and volatility evolution based on 2 correlated Wiener processes
o The model can be calibrated to suit the market data
American options – early exercise
Monte-Carlo Simulation – average random walk
Early exercise requires additional computation
o Solution of a system of linear equations
o No parallelizing Gauss method ~ O(n3)
o Parallelizing n multiplications and a number of additions