DSPy a system for AI to Write Prompts and Do Fine Tuning
Soa R 7 16 08 Appistry Private Clouds Etc Bob Lozano
1. Private Clouds: Cloud Computing for
Intelligence, Defense, & the Enterprise
“Fresh, radical and powerful
technology aimed at mee7ng the
needs of highly demanding business
problems.”
– Massimo Pezzini,
Gartner
The Fabric of Business www.appistry.com
2. Agenda
• Appistry Background
• Product and Posi9oning
• Case Study
• Cloud Compu9ng
The Fabric of Business www.appistry.com
3. Data-Driven World Driving
New Scalability Requirements
DATA INTENSE
(VOLUME)
• Data sources
• Sophis9cated analy9cs
• Internet distribu9on
• New business models
• System to System (SOA)
• Customer Expecta9ons
• Mobile Access
• Constant change
DEPARTMENTAL ENTERPRISE
APPLICATIONS
APPLICATIONS
CPU INTENSIVE
(# OF SERVERS)
The Fabric of Business www.appistry.com
4. Emerging Platform Requirements
Highly Capable:
• Very scalable
• Agile & adap9ve
• Reliable and fault‐tolerant
• PlaPorm independent
Yet Extremely Simple:
• Reduced development complexity
• Easier opera9ons
• Self‐managing
“Googliza9on for the masses”
The Fabric of Business www.appistry.com
5. Application Fabric Solution
Scale-out Application-level Automated
Virtualization Fault Tolerance Management
Fabric looks like a single Fabric provides software- Fabric & applications easy to
system: based reliability: deploy and manage
• Enables linear application • Application state • Dynamically discovers and
scale propagated to multiple assimilates new
machines at all times computers
• Complexity of distributed
software development • Ensures no job / transaction • Automatic provisioning of
hidden from developers ever fails software stack
• Complexity of distributed • Simplifies development and • Updates occur without
system management operations downtime
hidden from operators
• Enables use of commodity • Maximizes performance
hardware while minimizing resource
consumption
Applications ‘inherit’ these capabilities without manual coding
The Fabric of Business 5 www.appistry.com
6. Comprehensive Platform for
Scale-Out Applications
Appistry EAF
Policy Defini4on & Enforcement
Service Access
Fabric API Remo4ng
Web Services
Management & Monitoring
Deployment & Provisioning
Adap4ve/Distributed Load Balancing
Services
Java .NET C/C++ Legacy External
POJO PONO (Web)
Process‐Based Service
Orchestra4on
Data
Distributed Cache & State Management
Self‐Organiza4on
Linux
Windows …
x86
The Fabric of Business www.appistry.com
8. Case Study:
From Single-Threaded Algorithm in Lab to Large-Scale
Logistics Solution in Two Days
Appistry EAF at FedEx
In order to increase efficiencies, FedEx is
Data Process Result building a key logistics planning application on
Appistry EAF.
Challenges:
• Bringing strategic application to market quickly
• Supporting existing application code
• Ensuring predictable request execution
Results:
• Application deployed in one day
• Predictably processes all shipments in
required timeframe
• Application to save tens of millions of dollars
• Data: Database of today’s packages per years
• Process: Calculate routes for each truck based on
today’s packages
• Result: Calculated routes propagated to loading and
tracking systems
The Fabric of Business 8 www.appistry.com
9. Case Study - GeoEye
Appistry EAF at GeoEye
The leading provider of satellite imagery for
Data Process Result government and commercial applications,
GeoEye is building its next-generation image
processing applications on Appistry EAF.
Challenges:
• Multi-core / SMP development complexity
• Risk, cost and agility of traditional platforms
• Meeting customer SLAs
Results:
• Imaging applications now able to process in
excess of 5 TB of satellite imagery per day
• Developers able to focus on core competencies
• Capital savings greater than $1.2 million
• Data: Raw satellite image is retrieved from SAN and • Easily meet customer requirements for
broken up into 9les maximum processing time
• Process: Tiles are processed using proprietary GeoEye “ By relying on the application fabric to provide
algorithms for sharpening, geocorrec9on, etc. scalability, reliability and manageability, we can
leave our infrastructure concerns behind and focus
• Result: Tiles are reassembled and stored back in SAN on providing maximum value to our customers. ”
– Ray Helmering,
VP Photogrammetric Engineering at GeoEye
The Fabric of Business 9 www.appistry.com
10. Business Results / Value:
Demonstrated ability to:
• Decrease 9me‐to‐market
• Simplify development
• Reduce project risk
• Simplify opera9ons
• Provide scalability
• Enhance reliability
The Fabric of Business www.appistry.com
14. Why Cloud?
Enterprises want:
• U9lity acquisi9on
• Elas9c capacity (both up and down)
• Scale, Scale, Scale
• Improved quality‐of‐service
• Outsource to specialist provider
The Fabric of Business www.appistry.com
15. Cloud Providers
Migra2on to Value
SaaS
Applica9on‐Oriented
Force.com
Value Proposi9on EngineYard
PaaS
Google AppEngine
Joyent
Infrastructure
Amazon
Undifferen9ated HW, VMs Linode
The Fabric of Business www.appistry.com
17. Today’s Application Fabric &
Private Clouds
Private Cloud Private Cloud
Enterprise
The Fabric of Business www.appistry.com
18. Emerging Cloud Computing
Ecosystem
Public Virtual Private Cloud
Cloud
Private Cloud Private Cloud
Enterprise
The Fabric of Business www.appistry.com
19. Broader Cloud Opportunity
Appistry brings “applica9on awareness” to public, private,
virtual private clouds
Apps‐in‐a‐Cloud Cloud‐Enabled Applica9ons
• Move exis9ng apps to • Applica9on‐aware
u9lity‐priced 3rd party • Allows applica9ons to take
hos9ng advantage of Cloud
• Hopefully provider has • Elas9city
beher up9me than you do • Agility
Infrastructure Up Applica7on Down
The Fabric of Business www.appistry.com