Managed Grid Services and SaaS at COLT

Managed Grid Services and SaaS at COLTService Migration and the CGP

Frank FalconCOLT Managed Services

2

London

Brussels

Amsterdam Hamburg

Berlin

FrankfurtParis

Zuric

MilanTurin

BarcelonaMadrid

MAN with Data Centres

MAN

COLT connected city

Operational networkmanaged end-to-end

COLT today

> 15,000km European fibre network connecting 100+ cities

> Typically 96 fibres/route

> Latest 10Gbit/s optical technology

> 5,000km access networks in 32 European cities

> Industry-leading services

> First Ethernet service in 1993

> Ethernet over SDH from 2003

> Switched Ethernet services from 2005

> 14 data centres fully integrated with fibre network

> 60,000 sq metres data centre space

> Security Operations Centre

Ethernet SDH

IP/MPLS

DWDM

3

COLT evolution

Embracing Innovation within COLT

> Innovation is a fundamental driver for Colt as it evolves and it has embraced appropriate innovative developments to create a compelling company-wide Vision

Helping customers evolve

> Colt has recognised that Service Providers need to work closely with customers to help them with the transition from the IT-centric world of today to the web-centric world of tomorrow

4

Evolutionary drivers

1. Legacy applications

> Virtualisation of delivery capabilities at every level of the stack is driving Service migration

> Services are moving to “on demand”

> Migration path to the emerging Web-centric world

> Seamless and cost-effective

> Continuity of operations

> NO re-engineering of application software

2. Web-centric applications

> New applications from “mashing” plus ASP-type hosting

> Incorporate New and Legacy applications into delivery model

What are the challenges for each ……?

5

1. Legacy applications challenges

Within the Enterprise Data Centre environment

> Lack of agility due to statically-bound resources

> Applications are over-provisioned to peak loads

> Will not scale dynamically or adapt in any way

> Difficult to align resources with business imperatives

> Rising complexity and Total Cost of Ownership

> Server sprawl – physical and virtual

> Rising operational costs

> Distributed systems are costly to manage

> Increasing space, power, and cooling costs

> 5% - 7% CAGR floor space growth

> Increasing regulatory compliance requirements

> Current monitoring environments only deliver information rather than control

Test andDevelopment

StagingEnvironment

StagingEnvironment

High AvailabilityDR

StagingEnvironment

ProductInfrastructure



App A App B App C

6

1. Legacy applications challenges

Meeting the challenges needs:

1. Step-change improvement in operating costs

2. Future-proof environment for

continuing operation

Within the Enterprise Data Centre environment

> Lack of agility due to statically-bound resources

> Applications are over-provisioned to peak loads

> Will not scale dynamically or adapt in any way

> Difficult to align resources with business imperatives

> Rising complexity and Total Cost of Ownership

> Server sprawl – physical and virtual

> Rising operational costs

> Distributed systems are costly to manage

> Increasing space, power, and cooling costs

> 5% - 7% CAGR floor space growth

> Increasing regulatory compliance requirements

> Current monitoring environments only deliver information rather than control

7

1. Legacy applications – addressing the challenges

> Currently challenges addressed on an individual basis

> Hardware configurations and efficiencies upgrades

> Operational Process revisions

> New management/information systems

> Produces incremental gains but no “step-change”

> Deliver “step-change” improvement through holistic services

> Undertake three key actions

> Harness virtualisation

> Create Service Migration Path

> Create suite of services

> What does this model look like…?

8

1. Legacy applications - Services migration layers

Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute

Virtualised Application

Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers

Direct/Hosted Applications

Virtualised Applications

ST

AT

IC

ON

DE

MA

ND

Legacy Applications

Service Migration Path

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

9



Legacy Applications


Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute

Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers



ST

AT

IC

ON

DE

MA

ND ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND ST

AT

IC

ON

DE

MA

ND

10

Application virtualisation – NO code re-engineering

> Usually application virtualisation is partial in that the application software is allocated to a Cluster within which execution can be managed across the cluster resources

> Full application virtualisation is the complete separation of the application software from the hardware it executes upon. Run-time execution is directed across clusters by an intelligent device on a task-by-task basis

> Typically provided by wrapping the native application in a wrapper that specifies all the run-time parameters and submitting to the run-time intelligent controller

11

Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute


Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers



ST

AT

IC

ON

DE

MA

ND

Legacy Applications


ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND


What services will deliver the Step-change?

12

Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute


Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers



ST

AT

IC

ON

DE

MA

ND

UtilityCompute

Legacy Applications


ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND


13

Utility computing model

> A “dynamic-infrastructure” only play within Providers‘ environment

> Service Providers create capability for on-demand provisioning and sharing of their IT resource platform

> Service Providers offer Enterprise customers access to specified parts of this remote compute environment on pay-as-you-go basis

> Applications/tasks allocated onto infrastructure assets

> Sophisticated Load-balancing and Cluster management ensure performance parameters met by manipulating assets in pool

> Supports wide variety of real/virtual servers, OS, databases

> Customer benefits: NO software re-coding required; Low Capital expenditure; Peaks met without having to address own internal IT utilisation inefficiencies; cost-effective e.g. Sun Utility computing $1.00/cpu/hr; AT&T $0.35/cpu/hr

> Customer dis-benefits: No application-level guarantees; Must plan ahead and forecast needs accurately; Staff with expert team; Self-Manage operations; Still statically-bound at runtime

14

Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute


Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers



ST

AT

IC

ON

DE

MA

ND

UtilityCompute

GridCompute

Legacy Applications


ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND


15

Existing grid computing model

GRID Computing

> Application & dynamic-infrastructure play

> Grid engines incorporated within the Enterprise customer’s own LAN / Data Centre(s)

> Virtualises customers own IT assets raising utilisation rates from 5 ~ 10% to above 70%

AND

> Provides dynamic mapping of Users and Applications to resources (decouples jobs from underlying infrastructure)

> Applications/tasks allocated on a “pull” basis (polled)

> Customer benefits:

> NO software re-coding required;

> Substantially reduces TCO (40% reduction in hardware possible);

> Improves ROI by addressing utilisation inefficiencies

> Embryonic SLAs by Application performance management & flexibility controls

> Multiple simultaneous application-execution instances economically feasible

> Customer dis-benefits:

> Staff with expert team;

> Self-Manage operations;

> IT asset base limited by Firewall;

> Multi-node enterprise cannot utilise geographically-spread assets easily.

> Network connectivity still “statically” provided

16

Grids today & next steps

> Well-established model

> Many operational Enterprise Grids (some saturating)

> Provides application services to end-users

> Shares transparent resources across multiple users

> Operates to parameters set by Grid Administrator

Next step

> Provide explicit Application-level SLAs with QOS guarantees to users

Enterprise

WANS LANS

Grid Engine

17

Extending the grid concept – the CGP

> Create Grid platform with Network

> Add Network controls to virtualise connectivity – Adaptive Networks

> Use these to dynamically link to additional IT resources across Grids

> Add controls for the whole virtualised infrastructure through set of configurable policies

> Encapsulate control within a Policy Engine LINKED to Grid Engine

> Creates the COLT Grid Platform - the CGP

What does the CGP enable…?

Enterprise

WANS LANS

COLTNetwork

Networked Grid & Policy Engine

Existing Grid Engine

IPVPN

COLTIT Assets

18

Extending the grid concept – the CGP

CGP enables COLT to provide

> Managed Grid Services with explicitly specified Application-level SLAs

> Pro-active management of application performance within secure environment

And additionally

> Provides a future-proof Service Migration path…

> …through Managed Grid Services

Enterprise

WANS LANS

COLTNetwork

Networked Grid & Policy Engine

Existing Grid Engine

IPVPN

COLTIT Assets

19

Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute


Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers



ST

AT

IC

ON

DE

MA

ND

UtilityCompute

GridCompute

Managed Grid Services

Legacy Applications


ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND


20

Managed grid services model

Network-centric GRID Computing

> Application & dynamic-infrastructure including network connectivity play

> Grid engines incorporated within the COLT Network

> Virtualises customers own IT assets together with the COLT IT assets and network assets

> Adds a Policy Engine to the management layer

> Provides dynamic mapping of Users and Applications to resources under the control of the Policy Engine

> Applications/tasks allocated/moved anywhere into the Pool on a “push” basis by the Policy and Grid engines working in tandem

> Full application virtualisation with NO code re-engineering

> Provides compete end-to-end virtualisation as a Network-centric Managed Grid Service delivering complete end-to-end managements controls

> Pro-active management of applications within a highly secure environment

Customer benefits:

> Provides Application-Level SLAs;

> SLAs provided at differing levels mediated by customers QoS requirements against costs

> SLAs tailored to meet individual customer’s specific needs

> Staged buying/implementation decisions with simple fall-back at each stage

21

Full application virtualisation - Computational

Existing grid engines send executables and data to

capable grid agents in core

GRID Core

Under control of a policy engine grid engines send executables and data to specific grid agents across grid cores

Development

Submission

Registration

● Submits● Retrieves● Status

AgentAgent

AgentAgent

ApplicationDeveloper

NativeApplication

BuildModule

ApplicationUser Interface

GRID ServicesAPI’s

GRIDEngine

PolicyEngine

NativeApplication

Self-extracting XML wrapper

defining environment

AgentAgent

AgentAgent

> Policy Engine enables dynamic DARK-SERVER working> Pay ONLY when server is operational

22

Providing the step-change improvements

> Through the provision of Managed Grid Services COLT will enable customers to implement a step-change improvement in the operation of their existing legacy applications without costly re-engineering of application software…. as they migrate to the future service delivery models

> Services and the subsequent migration path will be tailored to meet the differing needs of customers

> Customers already running an internal Grid

> Customers ready to take the first steps on the migration path

> Customers needing to address specific application performance challenges

23

Typical managed grid services suite

Full managed grid service based on

application level SLA’s

Managed resilience grid service

Pay-as-you-go utility computing

New grid users Experienced grid users

Services

Customer-managednetwork-centric

grid service

Network-centricgrid extension

service

24

Typical managed grid services suite

> For existing Grid users a Grid extension service based on Managed Grid Computing services running on customer’s own virtualised IT assets plus Colt virtualised IT/Network assets

> For new customers a Managed Grid Computing service based on the COLT network-centric Grid Engine. Service initially managed by the customer himself running on that customer’s own virtualised IT assets

> For all customers a Utility Computing service on a pay-as-you-go basis, delivered through virtual servers on a variable costs model

> For all customers a Managed Resilience Service based on multiple site working and/or multiple application instances working reducing the effect of outages to near zero

> For all customers a Managed Grid Computing services managed by Colt running on sites within a “pool” of virtualised assets mediated by “Colt Value Optimiser” requirements. Provides Application-level SLA guarantees of various grades and prices for single and multiple application cases

25

Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute


Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers



ST

AT

IC

ON

DE

MA

ND

UtilityCompute

GridCompute


Legacy Applications


ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND


Future Services?

26

SaaS

2. Services migration layersWeb-based Applications

Tomorrows Business Driver

Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute


Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers



ST

AT

IC

ON

DE

MA

ND

UtilityCompute

GridCompute


ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

27

2. Web-centric challenges - SaaS

Trend

> One of the key trends in tomorrows computing environment will be Software as a Service (SaaS), where businesses no longer have to buy and maintain software applications themselves, but access them ‘on demand’ via the web(Google’s view is “tomorrow” has arrived …)

Challenges

> Requires complete separation of Applications and the Hardware Platform they run on

> With NO application code re-engineering

> Support Application development / mashing / hosting

> Support multi-tenancy and multi-users

> Provide virtual run-time space for applications

> Dynamic, configurable, configuration controls

SaaS

Web-based Applications


28

SaaS

Services migration layersWeb-based Applications


Network Layer

Storage Layer

Compute Layer

Application Layer

Virtualised Network

Virtualised Storage

Virtualised Compute


Private circuits

VPNWAN/LAN

Dedicated storage

Virtualised storage

Dedicated servers

Virtualservers



ST

AT

IC

ON

DE

MA

ND

UtilityCompute

GridCompute


Legacy Applications


ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

ST

AT

IC

ON

DE

MA

ND

API / SDK

GridPlatform

29

Meeting the challenge

COLT is focusing on developing innovative infrastructure and services, such as Managed Grid Services incorporating APIs / SDKs through which customers can seamlessly migrate to tomorrow’s web-based services

Typical service

SaaS offering based on the Colt Managed Grid Service providing the enabling infrastructure for ISVs and application providers to develop and host their application images with all the service controls necessary for multi-tenancy

SaaS

Web-based Applications


Virtualised Network

Virtualised Storage

Virtualised Compute


API / SDK

GridPlatform

30

> Progressively virtualise additional Enterprise’s ICT resources

> Link in logically to the resource pool

> Provide Application-Level SLA guarantees controlled by Policy Engine

CGP adoption strategy

> CGP Service initially virtualises a single Enterprise’s assets

> Links them logically with COLT’s own IT and Network connectivity assets

COLT Managed

IT Resource“Pool”

N/W Resource“Pool”

COLT Managed

COLTInternal IT Assets

Enterprise A

WANS LANS

Application“Virtual” runtime space

Applications

ServiceProvider

GRID Engine

Policy Controls

ResourceBroker/Trader

31

THE EVOLUTION OF COMPUTING TO GRID

COMPUTINGAS A UTILITY

Job submission

Mainframes (few)

1950’s – 1960’s

Job submission

GRID ENGINE

2000’s – 2010’s

Enabled by Service Providers with:-1. Adaptive High-Bandwidth IP

Network connectivity 2. Web-centric business applications

delivery capability

MANAGED SERVICES DELIVERED ON UTILITY

ENVIRONMENT ON A MASSIVE SCALE

Mini-Computers (Thousands)

Job submission

1960’s – 1970’s

PCs / Servers(Millions)

Job submission

1970’s – 1980’s

PDAs(10s Millions)

Job submission

1990’s – 2000’s

USER EFFICIENCY

PLATFORMUTILISATIONRATE

0

32

Next steps

COLT and selected vanguard customers drive developments

> Suite of services tailored to agreed requirements

> Create tailored service migration path capabilities

Thank You – Any Questions?

33

Adaptive networks

> The Adaptive Network, derived from software-controlled virtualised connectivity, is a key component of the new Network-Centric SaaS world

> COLT will provide Adaptive Networks by offering virtualised connectivity configuration controls at levels 1, 2 and 3 in the OSI stack to dynamically meet application performance requirements

> Network controls of the virtualised connectivity will include:

> Manipulation of MPLS VPN labels and Class of Service tags

> Dynamic control of router tables

> Flow control at the Physical and Data Link layers

Video

SAP

Email

Web Browsing

Voice

AFClasses

OverallBandwidthDE

Classes

EFClass

MPLS Classes of service

1

2

3

4

Managed Grid Services and SaaS at COLT

Documents

manage increasing

legacy applications

bound resources

grid concept

centric world

level slas

legacy applications

cooling costs