250 likes | 469 Views
The Organic Infrastructure. CRM. ERP. Financial. Portal. Document Mgmt. 5 Separate Web Farms5 Separate SQL Environments5 Separate Identity Stores. The Organic Infrastructure. CRM. ERP. Financial. Portal. Document Mgmt. IT PainSeparate Identity StoresSeparate and inconsistent SecuritySeparate Config and DeploymentSeparate Resilience/Load BalancingSeparate Monitoring and Management.
E N D
1. Rethinking Infrastructure Architecture: Service Oriented Infrastructure Kevin Sangwell
Infrastructure Architect
Microsoft EMEA HQ
2. The Organic Infrastructure Each new application or part of infrastructure is introduced as a stovepipe.Each new application or part of infrastructure is introduced as a stovepipe.
3. The Organic Infrastructure Higher cost, more complexity
Less secure (no end-to-end identity management)
Compliance difficult/expensive to enforce cross-company
CRM, ERP vendors are typically 3rd party. However these are now being forced to take an SOA perspective.Higher cost, more complexity
Less secure (no end-to-end identity management)
Compliance difficult/expensive to enforce cross-company
CRM, ERP vendors are typically 3rd party. However these are now being forced to take an SOA perspective.
4. The Organic Infrastructure Poor user experience re-enforces the “IT as a necessary evil” mindset
Lower productivity
These all increase business risk
Cost implications of diuplication
Poor service experience for customers (inconsistencies)Poor user experience re-enforces the “IT as a necessary evil” mindset
Lower productivity
These all increase business risk
Cost implications of diuplication
Poor service experience for customers (inconsistencies)
5. Consolidation is the answer, right? Reduces number of stove pipes, but doesn’t solve them
Next application/project adds another stovepipe
I think of this as “backwards consolidation”
Doesn’t change thinking The problem with consolidation from an SOI perspective is that it is viewed as a project. Once consolidation is complete, its finished. There is no change in approach for future projects, resulting in more stovepipes. SOI fundamentally changes the thinking from stovepipes and solutions to shared services which applications can consume.
The definition of insanity if doing the same thing and expecting different results!The problem with consolidation from an SOI perspective is that it is viewed as a project. Once consolidation is complete, its finished. There is no change in approach for future projects, resulting in more stovepipes. SOI fundamentally changes the thinking from stovepipes and solutions to shared services which applications can consume.
The definition of insanity if doing the same thing and expecting different results!
6. SOI: What it looks like
7. SOI: What it looks like
8. Getting There
9. Define & prioritise services according to ROI
Put low hanging fruit at the top
The difference between centralised and service-oriented is “shared service” ROI=Estimated cost savings (e.g. lower management overhead) and improved user experience vs technical complexity and scale of service
ROI=Estimated cost savings (e.g. lower management overhead) and improved user experience vs technical complexity and scale of service
10. Good candidates
Identity Management / Directory
Web Hosting
Database
File store
11. If IT infrastructure is obvious to the business = poor perception of IT
IT Infrastructure is not designed around users
Seek to improve Enterprise user experience
Unified view (network drive, published printers)
Single sign-on
Location independence/roaming
User Subscriber experience User experience is one of the key things which, if done correctly, will make the business think of IT as a strategic asset. The moment IT becomes “exposed” to the business (e.g. having to manually map network drives, call helpdesk to add a new printer when roaming), it becomes an overhead: poor perception, and costs productivity
At this phase, we’re looking at the enterprise user experience, not service user experience.
Subscriber experience = graceful failure mode: when service dependency fails, give users a “service currently unavailable” page rather than application errorUser experience is one of the key things which, if done correctly, will make the business think of IT as a strategic asset. The moment IT becomes “exposed” to the business (e.g. having to manually map network drives, call helpdesk to add a new printer when roaming), it becomes an overhead: poor perception, and costs productivity
At this phase, we’re looking at the enterprise user experience, not service user experience.
Subscriber experience = graceful failure mode: when service dependency fails, give users a “service currently unavailable” page rather than application error
12. Forward consolidation for each service
Attach to Projects
Major pain/cost areas such as IDM
13. Forward consolidation The future is difficult to predict - what i/o, RAM, CPU will my future application need … so
Abstract & Standardise
Categorise subscribers as High, Medium or Low
Capacity (storage & bandwidth)
Load (concurrency / transactions)
Performance (responsiveness / user expectations)
Availability
Implement Standard platform (hardware/software) for each of above
When you’re defining services in the application architecture domain (SOA) you should be doing this already. Infrastructure Architects can help Solution Architects in this areaInfrastructure Architects can help Solution Architects in this area
14. Backward consolidation
Low hanging fruit
Challenges
QOS: many services don’t support QOS
15. Assign Service Manager for each service
Owns relationship with other services
Subscribers
Publisher
Service Delivery
Service Level Management
Capacity Management
Availability Management
IT Continuity Management
Financial Management
Service Support Service Support
Incident management
Problem management
Configuration management
Change management
Release management
Service Support
Incident management
Problem management
Configuration management
Change management
Release management
16. Blockers Technology
Security
Regulatory & compliance
Aim to centralise these instead of service-orient them Technology:
no QOS.
Support of vendor (e.g. moving functionality out of stove pipe)
Performance: moving functionality off hosts (e.g. MIIS and SQL)Technology:
no QOS.
Support of vendor (e.g. moving functionality out of stove pipe)
Performance: moving functionality off hosts (e.g. MIIS and SQL)
17. SOI Enablers/facilitators Virtualisation is your friend, and your enemy
But doesn’t solve all problems: remember virtual hosts still need managing & are lower performance
Clustering
Cost of resilience reduces with addition of services
SAN
Flexibility; capacity, replication, backup
Evaluate on a case-by-case
Slower than DAS
Some applications don’t support SAN replication/backup Virtualisation:
some QOS controls
isolation (security/regulatory)
Clustering:
Offering better SLAs for minimal extra cost
Lowering operational cost – maintenance easier to perform in service hours
Virtualisation:
some QOS controls
isolation (security/regulatory)
Clustering:
Offering better SLAs for minimal extra cost
Lowering operational cost – maintenance easier to perform in service hours
18. An Example: File & Print Easy for central sites: clusters
Difficult for branch:
Either centralise & cache or replicate data
Migrate users to DFS for its location-awareness & server abstraction
Fileserver Migration Toolkit does the legwork
Don’t forget this is a service; apps can consume
Benefits
Reduces backup pain
Easier to manage/apply policies/report on
Easy to add resilience & DR support
19. Example: Identity Management Service Define Service:
Single directory of users for authentication and access control
User Experience
Transparency (SSO, location independence, discoverability)
Subscriber Experience (Capabilities)
LDAP Directory (e.g. AD)
Authentication (LDAP Bind, NTLM, Kerberos)
Authorisation (Group membership)
Auditing (directory access)
20. Example: Identity and Access Management Design Logical Service
Capacity
Performance
Scalability
Backup & DR
Security
Extensibility for subscribers
Design Physical Service
Server sizes
Server locations Capacity: how much storage? How much network bandwidth
Performance: user/subscriber load
Scalability model: up, out, partition?Capacity: how much storage? How much network bandwidth
Performance: user/subscriber load
Scalability model: up, out, partition?
21. Extensibility Remember “blockers”?
Technology (Schema)
Regulatory (Forest)
Security (Account Policies) Schema: use ADAM as application directory where you can extend the schema. Proxy calls into AD meaning you still take advantage of the directory “service”.
Regulatory: split the users across forests (or maybe domains). Establish trusts if possible.
Security: split the users across domainsSchema: use ADAM as application directory where you can extend the schema. Proxy calls into AD meaning you still take advantage of the directory “service”.
Regulatory: split the users across forests (or maybe domains). Establish trusts if possible.
Security: split the users across domains
22. Example: Identity and Access Management Service Evolution
Move to Identity Management Service
Provisioning/de-Provisioning triggered from HR database
Federation
User Self Service
All subscribers benefit from these capabilities
23. Does SOI really have an ROI Gartner modelGartner model
24. Architecture Design Review 1 Day engagement
Follow-up report
Limited number
25. Question & Answer Panel