ServiceMesh's Dave Roberts presented, originally at Cloud Connect 2012, Santa Clara CA
Applications are the heart of enterprise IT. Everything that IT does is focused on delivering new functionality through applications. Unfortunately, for most enterprises, the application delivery process is tremendously inefficient, leading outcomes where more application development projects are canceled than are completed.
Tampa BSides - Chef's Tour of Microsoft Security Adoption Framework (SAF)
Applications at Scale
1. THE AGILE IT PLATFORM
Applications at Scale
Dave Roberts
SVP, Business Development and Platform Ecosystem
dave.roberts@servicemesh.com
Twitter: @sandhillstrat
2. ServiceMesh Overview
Continuous Big-Data Analytics
Enterprise PaaS
Delivery (e.g. Grid, Hadoop)
Governance
Business Policies Policy- Complex
Auto-
scaling,
Version-
Self-service controlled
controlled topology elasticity,
App Store asset
API provisioning and SLA
repository
Workflow mgmt.
Access Management
ServiceMesh
Service Management Agility Platform
Resource Management
Public and Private Cloud Implementations
Resources
2
13. “Is this application suitable for the cloud?”
• Cloud Value refers to economic, High-priority
candidates
business, and functional filters
that assign significant business
value when transitioned to cloud
environments.
Cloud Readiness
• Cloud Readiness refers to
characteristics of latency,
geography, responsiveness,
platform, data, and organizational
requirements that lend itself to
cloud environments.
Cloud Value
Low-priority
candidates
13
14. Migration Analysis Categories
Application
Portfolio
Migration Group 1 Migration Group 2 Migration Group 3
Layer “Stateful” “Less Stateful” “Stateless /
Cloud Native”
Business Logic Unchanged – embedded in Unchanged Refactored to align with “cloud
stateful image native” design patterns
Application Framework Unchanged – embedded in Possible upgrade to the Refactored to align with “cloud
stateful image current IPaaS Platform & OS native” design patterns
versions
Platform (e.g. IIS/.Net) Unchanged – embedded in Possible version upgrade Possible version upgrade
stateful image
Operating System (e.g. Unchanged – embedded in Possible version upgrade Possible version upgrade
Windows 2008) stateful image
Agility Integration Yes Yes Yes
14
15. 1 Planning at Scale
2 Construction at Scale
3 Deployment and Operations at Scale
15
17. Provide standardized “as-a-service” building blocks to lower
support costs, risks, and improve productivity
Design Development Test Operations
Self-service portal
Unified governance
and visibility
Portfolio of Role-based access
standardized
offerings:
iPaaS Offerings Standardized Packages 3rd-Party PaaS/SaaS Enterprise PaaS
Windows 2008 Monitoring agents Salesforce.com Hadoop
RHEL HIDS Force.com Cloud Foundry
SQL Server Anti-virus Azure Apprenda
IIS.NET Virtual Firewalls Google Apps WebSphere
others… others… others… WebLogic
17
18. Leverage existing tools to access resources while
retaining full control
Design Development Test Operations
SDLC systems others…
and tools:
API access REST-based API
Unified governance
and visibility
Role-based access
control via API
Bare Metal
Dev Clouds QA Clouds others…
Environments
1-off sandbox
environments
18
19. The Environment is an Artifact!
Environment = VM topology, middleware configs, firewall rules, etc.
Development creates Dev QA Prod
application runtime Dev Cloud Reqmnts QA Cloud Reqmnts Prod Cloud Reqmnts
environment as part of
the development
process.
The environment travels
with the code and data
sets as a part of the
development process. Prod Cloud
Dev Cloud QA Cloud
What gets tested and Internal Private Cloud
runs in production is
exactly what
development created.
Many other providers…
19
20. 1 Planning at Scale
2 Construction at Scale
3 Deployment and Operations at Scale
20
21. 1. Agile Ops
2. Traceability
3. Policy and Governance
4. Reporting and SLA Management
21
22. Characteristics of Agile Operations
Traditional Agile Ops
Dev Deployed
App 1
Dev Deployed
App 2 Apps 1…n
App 3
Deployment takes weeks or months Deployment takes minutes
Manual, custom assembly for each app Automated, repeatable outcomes
Craftsman model with high labor costs that scale
Craftsman build the robots instead of
working in the assembly line
22
23. Reference architecture for traceability and control
Deliver precise versioning, roll-back and deployment management.
Monitoring/PerfMgmt
Chargeback/Billing
LDAP/AD
Services and Stateless Environment
Others Resources Images Configuration
Code
Development
Base
Environment
Dev Team
Deploy
Code Test
Base Environment
Repository Mgr Deploy
Dev Team • Images
• Packages Deploy
Code • Resources Staging
• Configuration Environment
Base Deploy
Dev Team
Production
Code Environment
Base
Dev Team
Development Operations
23
24. Enforce fine-grain policies to govern across teams,
resources, and environments
London External NYC Dev India QA
Dev Team Consultants Team Team
Sample Policies:
Regulatory: Live test data
can only be deployed on
secure private cloud.
Create and enforce policies
Access: Contractors are End-to-end audit trail
permitted access only to Extensible policy framework
external public cloud Policies control
access to resources
accounts.
Policies
determine right-
placement
24
26. Manage SLAs with automatic thresholds,
compound rules, and custom notifications
Setup highly detailed performance
Create compound rules thresholds for auto-scaling
based on chains of events
Create custom notifications based on
different lifecycle events
26
28. Customer Example: Time-to-Market Reduction
Current External Cloud Internal
Baseline Provider Pre-Production Cloud
Platform 70 days Not Available -52 days (41% reduction)
Client specific
patches/utilities
17 days -11 days (-8%) -11 days (8% reduction)
IaaS (Base OS) 40 days -36 days (-29%) -36 days (29% reduction)
Total
Provisioning 127 days -47 days (37%) -99 days (78% reduction)
Time
NOTE: All time-to-market scenarios include a conservative 20-days
allocated for general purpose build and test tasks.
28
30. Thank you
Dave Roberts
Vice President, Business Development and Platform Ecosystem
Email: dave.roberts@servicemesh.com
Web: http://www.servicemesh.com/
Company Twitter: @servicemesh
Personal Twitter: @sandhillstrat
30
Editor's Notes
NOTES: Earlier I mentioned that Continuous Delivery encompasses developer resources and deployment environments… not just the code. Talk a little about the “resources” side of that statement.
NOTES: Its more than just self-service portals….Folks want to access deployment environments seamlessly as part of their daily work routine. You want to access it from the tools you already use, and in most cases you don’t need to be exposed to the inner workings of how those environments are getting spun up. If you can provide these on-demand and tailored for the user’s needs… you can eliminate significant wait time. Agility has a REST-based API that is full featured. That includes role-based access rights managed through the API. You can access Agility from a wide range of tools to spin up an unlimited variety of deployment environments…. All of which are fully governed and controlled. These environments are all tracked and versioned as well… so you can share them with other teams, and/or roll back as needed.
Center = artifact repository. Do this well = version everything. Precise versioning.Product apps…on the left side.Architect producing platforms and infrastructure (ServiceMesh)Configuration has 2 faces: environment independent. Environment specific configuration.Problem? Probably related to a software change. Ask “what changed”? Need a yesterday button with precisely versioned artifacts. (or don’t repair…reprovision).Strong versioning. This scales. A lot of participation of organization to publish change. Apache makes a change. Not everyone will push the update right away. Not on same day. Rolling upgrades instead. Have to have that kind of infrastructure. Zooming into the 80% problem. This is future state. Current state is series of tailored applications. Each LOB has a separate process. No economies of scale. A bunch of small independent units with different processes. Each group on left going to a separate environment in parallel. This x number of LOBs. (Stop 32:09)
Enterprise organizations can have hundreds of separate pre-production teams…. All of which need access to resources and environments. Providing automated, policy-driven governance is important to efficiently empower these teams with self-service access to the resource (such as platforms and applications) that they need, along with rightplacement to the appropriate sandbox environments. Agility Platform allows you to create and enforce these policies, which are leveraged across the SDLC to give you an end-to-end audit trail. The policy framework is also designed for extensibility… so that policies can be easily created for unique industry or organizational needs. For example… you could create a regulatory policy… such as a requirement that Live test data can only be used on the secure private cloud.….. Or a security policy…. So that external dev contractors can only use external public cloud accounts.
NOTES: (OSI/SOA refers to the client’s approved Operating System Image / Standard Operation Environment, which includes a range of the client’s patches and client specific utilities. )Traditional data center: this current approach includes 20 days for customer application build and test processes. It is assumed this is consistent across all other approaches as well. Pre-Production Cloud:Reductions mainly in automated provisioning and for design work across all layers.External Cloud Provider (Fujitsu and HP)CSPs were asked primarily for data on their core IaaS services. “Platform” services were not available from the CSPs. OSI/SOE are assumed to be the same as for pre-production cloud, but are likely to be less as an increase in engagement effort and decision gateways would be likely.