Rai Technology Technology Overview October 2009 About Rai Founded - - PowerPoint PPT Presentation

Rai Technology

Technology Overview October 2009

About Rai

• Founded 2003 in San Francisco, California.
• Offices in USA, UK and Japan.
• 20+ customers including many top tier financial

institutions around the globe.

• Profitable & stable.
• Support our customers in all time zones 365 x

24 x7.

Our Technologies

• Middleware Monitoring:
• Rai Insight
• Market Data Distribution Service:
• Rai Cache
• Rai ICE
• Rai API

Middleware Monitoring

• Rai Insight:
• Network pcap capture.
• Decodes TIBCO Rendezvous, ciServer and 29West LBM packets.
• Continuous real-time monitoring.
• Rich UI with real time graphs and statistics.
• Round trip latency analysis.
• Incident analysis.
• Usage statistics.
• Alerting capability.
• Reporting.

Market Data Distribution

• Rai Cache
• Rai ICE
• Rai API

Generic Market Data System

Market Data Service Provider Publishing Service Consumer API Publishing Service Consumer API Subscribing Service Consumer API Pub & Sub Service Consumer API Subscribing Service Consumer API Subscribing Service Consumer API Subscribing Service Consumer API Network Protocol Asynchronous Message Flow

Vendor Lock-in

Vendor A Service Provider Service Consumer Vendor A API Vendor A Protocol Vendor B Service Provider Service Consumer Vendor B API Vendor B Protocol Vendor C Service Provider Service Consumer Vendor C API Vendor C Protocol

Rai Market Data Distribution System

Service Consumer 4 RAI API Subscribing Service Consumer API Subscribing Service Consumer API Service Consumer 5 RAI API Service Consumer 1 Vendor A API Vendor A Protocol Service Consumer 2 Vendor B API Vendor B Protocol Service Consumer 3 Vendor C API Vendor C Protocol Service Consumer 6 RAI API RAI Open Protocol Service Consumer 8 RAI API Any Future Protocol Service Consumer 9 Future API Embedded Consumer 7 RAI Multi Protocol Market Data Delivery System RAI API

Use Case: TIB Replacement

TIC CI CI CI RVD RVD RVD App App App App RAI Cache ciRV RVD ciRV RVD RVD RVD App App App App QForms

• ver SASS

TIBMsg, QForms or Marketfeed

• ver SASS2
• r SASS3

Legacy Deployment RAI Drop-in Replacement CI Legacy TIBCO ciServer, the predecessor to RVD Legacy TIBCO Information Cache, a last value cache service. TIC RVD TIBCO Rendezvous Daemon. App An application written against one of the legacy TIBCO APIs RAI Cache The core component

• f RAIs multi-protocol

market data system. ciRV Bridge legacy ciServer clients to Rendezvous KEY RVD RVD

Use Case: High Low Speed Bridge

Wombat over 29West LBM Wombat FH Wombat FH Algo Engine W

• mbat

Protocol SASS3 Protocol Algo Engine Store ICE RMDS GUI GUI GUI High bandwidth, low latency Lower bandwidth, not highly latency sensitive. KEY Wombat FH Wombat Feed Handler. RAI Cache Internal Modules Algo Engine Algorithm trading engine consuming Wombat market service and contributed calculated data. Wombat Protocol RAI Cache protocol for handling Wombat market data system. Store RAI Cache internal message store. ICE RAI Cache Intelligent Conflation Engine. SASS3 Protocol RAI Cache protocol for SASS3 over RV . GUI Non latency sensitive GUI consumers of SASS3 market data system.

Use Case: Low Latency Solution

Store CAPR Protocol Embedded API/Proto Embedded FH RAI API RMDA TX Embedded FH RAI API Low Latency Subscriber RAI API RAI Cache Proces s RAI API RMDA TX Infiniband Network End User Processes KEY Embedded FH Feed handler developed on RAI API, built and deployed as a embedded service within the RAI Cache. RAI API RAI API interface Embedded API/Proto Implementation of RAI API for communicating with in-process instance of RAI Cache. CAPR Protocol Cache Protocol, service provider side. RMDA TX RMDA over Infiniband transport implementation CAPR API Low Latency Subscriber RAI API RMDA TX CAPR API CAPR API API implementation. Cache Protocol, service consumer side . Low Latency Subscriber Market data subscriber developed on RAI API, built and deployed as standalone application.

Rai MDDS Components

• Rai Cache:
• Protocol neutral/extensible.
• Subscription Management.
• Last Value Cache.
• Full image and delta updates.
• Dictionary download services.
• Conflation.
• Delayed Data.
• Optimized file based backing store.
• Fault tolerance.
• Arbitrary horizontal scaling.
• HTTP administration.
• Rai API:
• Protocol neutral/extensible.
• Access all Rai Cache services.
• C++, Java & C#
• Rai ICE
• Intelligent Conflation Engine.
• Each outgoing channel in the Cache can

set its conflation parameters independently.

• Alternatively, can be deployed a separate

process to the Cache.

• Can configure limit as messages and/or

bytes per second.

Rai Cache Architecture

SubjectMessage A.A.A A.A.B … … Z.Z.Z 0x___ 0x___ … … 0x___ Subscription Sources Broadcast Sources Interactive Sources Protocols In-Memory Store Services RAI Cache Service Consumers

• Subject: String name logical data channel.

Standard dotted notation.

• Message: A buffer of data packed according to

some scheme.

• Service: Internal client of in-memory store.

Generally used to build ancilliary services, such as the backing store.

• Protocol: Internal client of in-memory and

implementation of server side of a specific market data service protocol.

• Source: In or out channels covering some

configurable root subject.

• Broadcast: Incoming broadcast data.
• Subscription: Handles client subscriptions.
• Interactive: Like a broadcast source but receives

notifications of subscription start/stops.

Services and Protocols

• Protocols:
• SASS over CI
• SASS2/3 over RV
• Wombat MAMA
• Reuters RFA 5.x &

SSL

• Cache Protocol

(CAPR)

• Services:
• Store
• Delay
• Admin

Rai Protocols

• Cache Protocol

(CAPR)

• Rai designed

transport and payload agnostic protocol.

• Rai Cache supports

dynamic loading of transports and message formats.

• Full specification
• penly published.
• Embedded “Protocol”:
• Cache client and

cache in same process space.

• Cache client uses

RaiAPI

• RaiAPI

implementation communicates directly with in-process Cache.

Rai API

• Mostly pure interface, some abstract classes.
• Runtime implementation required for each

specific market data service protocol. Currently implementations for:

• All classic TIB protocols.
• The Cache Protocol.
• Embedded Protocol.

• http://www.raitechnology.com
• info@raitechnology.com