Date post: | 14-Jun-2015 |
Category: |
Business |
Upload: | davide-zari |
View: | 132 times |
Download: | 2 times |
Copyright © 2010 Parallel Trading Systems Srl
Parallel Trading Systems
Document prepared by : Davide Zari
October 2010
Business proposition
2/9
Team
MEMBERS EXPERIENCE
Global executive, 18 years in EMC launching Bil$ new BUs within international assignments. He lead two startups now acquired by Fortune 500 companies.
BSc in International Business at Boston College, executive programs
Marco Buchbinder
Senior Trader, equity derivatives market maker for Archelon Deutschland and Abaxbank.
MSc in Economics at Torino University.
Massimo Gallizio
More than 10 years of experience on front-office system development for Morgan Stanley, Credit Suisse, UBS, Paloma Partners, Citibank.
MSc in Software Engineering at Ecole Superieure en Sciences Informatiques, Sophia Antipolis, France.
Alberto Giannetti
Trading system development in ION Trading and Enel. Specialized on grid computing and high performance systems.
BSc in Computer Science at Pisa University.
Luigi Idà
VP of Business Development in many eComm startups, he lead multi-Mil$ sales programs to grow revenue and stock market value.
BSc at Boston College, MSc at Duke University Fuqua School of Business.
Chris Keyes
Managed Mil$ projects for Banca d’Italia, Credit Suisse Italy, Banca IntesaSanPaolo. Business Development for foreign markets.
EMBA at MIP-Politecnico di Milano, MSc in Physics at Trento University.
Davide Zari
ADVISOR
ADVISOR
3/9
Partners for growth
PARTNERS PROFILE
PARTNER’S LOCATION
E4 (Reggio Emilia, Italy) produces HPC cluster servers and storage for the financial and scientific industries. Biggest processor customer distributor of Intel in South Europe. Supplier of Ferrari Racing and CERN.
Caspur (Rome, Italy) is a inter-university consortium that provide both state of the art HPC infrastructures and advisory services on quantitative finance parallel computing.
TA (Brugg, Switzerland) offers business support and office infrastructure, moderate taxation, qualified staff recruitment at 20 min from Zurich.
Mellanox Technologies (Yokneam, Israel) is a world leader supplier of connectivity solutions that optimize data center performance. Mellanox products deliver market-leading bandwidth, performance, scalability, power conservation and cost-effectiveness.
4/9
Business summary
Vision: use of parallel computing to achieve fast response to market changes on extended portfolio data sets.
Applications: direct market connectivity, algotrading and pricing systems, risk management.
Customers: statarb groups, inv funds,
investment banks, in USA and EU.
Target revenue: 7Mil€ after 5 years
Patent: patent granted from USPTO (08/2010).
Prototype: patenting technology demo
available.
External resources: external advisors for risk management and quantitative finance.
HIGH-FREQUENCY TRADING AND RISK MANAGEMENT
5/9
Market needs
Reliability of front-office systems during markets major stress periods.
Exponential growth of market data. Growth of automated high frequency
algorithmic trading.
Computing time for risk management.
End of Moore law. Multi-core architectures efficency. Newtwork architectures complexity.
Integration costs.
1
PROBLEMS
2
3
4
5
6
1. High Performance Computing technology based on stable solutions, result of decades of research in parallel applications: weather forecast and fluidodynamics.
1. Real-time system, able to parallelize the computing through standard interfaces between nodes. Easy integration with existing applications.
1. Immediate scalability following market
changes.
1. Licensing based on number of financial
instruments.
1. Re-use of available hardware.
PARALLEL TS SOLUTIONS
1
2
3
4
5
6/9
Market trends
IBM acquires InfoDyne for marketdata distribution (29/4/2008).
NYSE Euronext acquires Lombard Risk
Management (18/2/2008).
Sun Guard acquires GL Trade for front-office
and connectivity applications (4/8/2008).
Knight Capital signs agreement with Quant House for dark pool management (12/1/2009).
Co-location, installation of dedicated servers in market data-centers. New market distribution model.
Electronic market fragmentation: book modules aggregation to have a better liquidity management.
Exponential growth in the number of transactions. New contracts from CME and ICE.
Multi-assets real-time risk management. Improvement of management models and roles.
MERGERS & ACQUISITIONS - PARTNERSHIPS FUNCTIONAL TRENDS
7/9
Prototype
B&S Merton model pricing 5000 simulated underlyings on a 4 node 32 cores cluster
360,000 priced
instruments calculated in real-time (NPVs + greeks)
More than 10,000
market update/sec
More than 50,000 market events/sec (quotes+trades)
Tested scalability
from 1 to 4 nodes and 2 to 32 cores
ARCHITECTURE
Simulated market data
8/9
-80%
-70%
-60%
-50%
-40%
-30%
-20%
-10%
0%
500 1000 2000 5000 10000
Prototype benchmark & TCO
TCO: on a 3-year perspective there is a cost
advantage managing a large range of financial instruments.
Model: hw costs, installation and configuration, patching, licensing, maintenance respect a average front-office installation are taken into consideration in this model
TCO
N° instruments ΔTCO%
0
200
400
600
800
1000
1200
1400
1600
1800
1000 5000 10000
16IB
8IB
16TCP
8TCP
BENCHMARK
N° trade events/sec
ms
Benchmark: latency peak plotted for
massive greeks calculation deriving from a discount rate change in the marlet, using 8 or 16 core, TCP or Infiniband network and changing trade events number.
Results: scalability regardless the number of core used, advantage on the use of Infiniband.
9/9
Product and revenue structure
Market Data System
Proprietary Middleware
Order Mng. Risk Mng
Execution
Project wbs: already available proprietary middleware for task parallelization
I° MDS: direct market access and basic analytics to generate trade signals. II° Order Management: algotrading and program trading capabilities.
III° Risk Management: realtime risk and performance analysis such as VaR, S, Alpha calculation. IV° Execution: smart order routing, FIX integration.
Revenue structure: License: recurring license, no maintenance. Professional services: for customer personalization.
PRODUCT LOGICAL MODULES
Copyright © 2010 Parallel Trading Systems Srl
ANNEXES
Market trends
Algorithmic trading on total transactions (*)
(*) NB: Hedge Funds Review news 25-feb-2009: il 60% average volume US equities made from high frequency trading community (source: Aite Group)
-
20.000
40.000
60.000
80.000
100.000
120.000
140.000
160.000
180.000
mar-07
mag
-07
lug-07
set-0
7
nov-07
gen-08
mar-08
mag
-08
lug-08
set-0
8
nov-08
gen-09
mar-09
NYSE volume [Mil] IVX volatility index
Exponential growth of market data
Growth of automated high frequency trading (now 60% of total transactions)
Growing transaction volume (+70% last year)
Volatility growth (+100% last year)
MTF growth (BATS, Chi-X, etc.)
1. Improved scalability on analysis and execution
2. Improved data management
3. Integrated real-time risk management
TRENDS BUSINESS NEED
1 2 3
1
2
3
Technology trends
Growing n° of CPU core: high performance systems are using hundered or thousand of CPUs, with a growing core number for every CPU (12 core within 2010).
Latency decrease: all the MTS interfaces are reducing market operation latencies, coming in the sub-millisecond region.
High speed networks: available high-speed low latency server to server networks, proprietary (i.e. 20-40 Gb/s Mellanox Infiniband) or standard (10 Gb/s Ethernet).
1 2
3
BATS Trading latency measures
N° CPU Top500.org supercomputing systems
Mellanox Infiniband bandwidth evolution
1
2
3
Value proposition
Server
CO
MPETIT
ORS
PARALLEL
Server Server Server
Server Server Server
Server Server
Server
Data
conso
lidatio
n
Less servers
Less computational time
Scalability
Update/s
Markets
Instrument N°
1k 10k 100k 1k
10k
100k
= Market opportunity
Update/sec
Markets
Instrument N°
1k 10k 100k 1k
10k
Global trading portfolio
Better Sharpe ratio
100k
Single
server
portfolio
Data
conso
lidatio
n
Portfolio
hedging
Consolidated portfolio
Global portfolio
Global portfolio
Unique portfolio for risk balancing
Real-time risk management
Better Value ar Risk (VaR)
Single server
portfolio
COSTS REVENUE RISK
Server Server
Server Server
Cluster
T1
T2
T3
Single
server
portfolio
Single
server
portfolio
T1
Portfolio
hedging
Portfolio
hedging
Portfolio
hedging
Positioning
FU
NCTIO
NAL V
IEW
TECH
NO
LO
GY V
IEW
Multithreading Stream processing Message passing
Pro: easy update for existing applications
Cons: scalability limited by shared memory solutions
Pro: parallelization not limited to shared memory systems
Cons: performance limitations on double precision math, specific hw implementation
Pro: maximum parallelism granularity, hw independency
Cons: difficult implementation, no real-time tradition
Markets Market Data Systems Front office Middle office Back office
Prototype
I° step
Prototype
I° step
II° step
Distinctive characteristics
FUNCTIONAL ADVANTAGES
TECHNOLOGY ADVANTAGES
Core/soket/servers scalability, not only on core/socket
Further speed increase given by the use of proprietary Mellanox Infiniband network
Scalability on increasing the number of market update
Real-time portfolio update for market hedging
Real-time option pricing for a really big portfolio, regardless the market updates number
Competitive TCO respect other solutions
Realtime Batch
Market data
Risk management
Pricing, OM, Execution
POSITIONING
Situation and prizes
Patent request made at the United States Patent and Trademark Office (USPTO). Patent granted in August 2010.
The patents will allow a better investment protection and facilitate growth in target markets.
PATENT
2010Q4 2011Q1 2011Q2 2011Q3 2012Q1
Prototype completed
USA and Italian company branches incorporated
Agreements completed with tech partners
Agreements with CASPUR and Technopark Aargau.
Alpha-test currently in progress
Prototype completed
Company incorporated
Framework base Market data OM
Partners selection
Alfa testers Sell mode
EVENTS Winner
Finalist
COMPLETED STEPS
Finalist
Past
Futu
re
Partners - Mellanox
COMPANY
CUSTOMERS (FINANCE)
PRODUCTS
POSITIONING (*)
Background: Founded in 1999 within Sequoia Capital investment. Focused on datacenter semiconductor development using Next Generation I/O standard. $107Mil sales FY’08 (+28%).
Markets: Solutions in transversal markets Enterprise Data Center, High Performance Computing, Embedded, with enterpise applications within Automotive, Energy and finance vertical markets.
Customers/Countries: total 100+ customers. Offices in USA, Canada, Europa, Asia and India.
(*) Top500.org world supercomputers ranking for computing power, filtered on Mellanox Infiniband network fabric
ConnectX: Chip and adapters for high speed connections 10/20/40 Gb/s, virtualization.
InfiniScale IV: Datacenter systems and switch 36 port 40 Gb/s with advanced traffic management, ultra-low latency (1 μs).
AWARDS
….