fujitsu world tour 2017: hyper converged data centre

0 Copyright 2017 FUJITSU

Hyper-Converged – The Only Way to the Software-Defined Data Center?

Fujitsu World Tour 2017

#FujitsuWT


Hyper-Convergence – The Only Way to SDDC? A Reality Check of Hyper-Converged Infrastructures

Gernot Fels

Head of Integrated Systems International Product Marketing, Fujitsu


Impact of digitalization on DC infrastructures

Digital world

More data

Collect

Transport

Process

Store

Needs: Fast and flexible scalability, resilience, simplicity, efficiency.

Digital world More data More challenges to IT

Network Storage Servers


Traditional DC infrastructure

Network

Storage

Server Server Server

App App App

Compute Compute Compute

HW defined

Rigid

Manual ops

Silos

Slow

Complex, inflexible, expensive and skills demanding.


How to Change the Game?


Hyper-converged infrastructure

Characteristics

Interconnected standard servers

Local DAS, no external storage

Virtual instead of physical SAN

Data services (e.g. replication, deduplication)

HA built-in

Unified central management

Scale-out (performance, capacity)

Auto-discovery

SW-defined

Network

Software-Defined Compute

Software-Defined Storage


App

Compute

Disks

App

Compute

Disks

App

Compute

Disks

Virt

ual

SA

N



Characteristics





HA built-in



Auto-discovery

SW-defined

Benefits

Less components, less space

Less energy and cooling

Simplicity (less admin efforts)

Lower skills (lower-paid staff)

Start small and grow as you go

Flexible alignment to business

Great responsiveness

Business continuity

Lower TCO

Network




App

Compute

Disks

App

Compute

Disks

App

Compute

Disks

Virt

ual

SA

N



Characteristics





HA built-in



Auto-discovery

SW-defined

Benefits

Less components, less space

Less energy and cooling

Simplicity (less admin efforts)

Lower skills (lower-paid staff)

Start small and grow as you go

Flexible alignment to business

Great responsiveness

Business continuity

Lower TCO

Network




App

Compute

Disks

App

Compute

Disks

App

Compute

Disks

Virt

ual

SA

N

Hyper-converged – Silver bullet of IT?


Architectures in comparison

Classical Servers, LAN, storage, SAN

LCM for multiple components (different LC)

2 separate networks to maintain

HW-defined

Scalability by adding multiple components, primarily separately

Storage activities are separated from servers

Established for many years

Hyper-converged Servers (storage included), LAN

LCM for less components

More powerful LAN required

SW-defined

Scalability by adding servers, primarily linearly (compute & storage)

Servers are involved in storage activities

To be considered when sizing compute resources

Entered the market some 3 years ago


What do these differences mean for you?

Classical To be used for general purposes

Physical, virtual, multi-hypervisor

HW dependence, primarily related to storage

Well-known SW licensing (e.g. DB)

Higher operational complexity

Possibly higher HW cost, but lower SW cost

Mainly well established vendors

Hyper-converged Less usable for general purposes

Virtualization is quasi a must

SW dependence (related to virtualization SW)

Commercial restrictions (e.g. DB licensing)

Lower operational complexity

Possibly higher SW cost, but lower HW cost

Established vendors and newcomers


What do these differences mean for you?

Classical To be used for general purposes

Physical, virtual, multi-hypervisor

HW dependence, primarily related to storage

Well-known SW licensing (e.g. DB)

Higher operational complexity

Possibly higher HW cost, but lower SW cost

Primarily well established vendors

Hyper-converged Less usable for general purposes

Virtualization is quasi a must

SW dependence (related to virtualization SW)

Commercial restrictions (e.g. DB licensing)

Lower operational complexity

Possibly higher SW cost, but lower HW cost

Established vendors and newcomers

Both approaches provide unique advantages. Good reasons for coexistence.


Classical or hyper-converged? – Aspects to be considered Workloads need scale-out / scale-up?

Compute & storage needs scale in tandem?

Granular expansion on component level?

Expected growth

ROBO

Mixed operation

Multi-hypervisor, bare-metal

Workloads benefit from data services?

Benefit from unified management?

Existing skills, staff roles

Utilize or replace existing storage

Storage capacity

SW licensing

CAPEX and OPEX


Classical or hyper-converged? – Aspects to be considered Workloads need scale-out / scale-up?

Compute & storage needs scale in tandem?

Granular expansion on component level?

Expected growth

ROBO

Mixed operation

Multi-hypervisor, bare-metal

Workloads benefit from data services?

Benefit from unified management?

Existing skills, staff roles

Utilize or replace existing storage

Storage capacity

SW licensing

CAPEX and OPEX

Use cases matter. Business requirements determine the optimum approach.

Recommendation: Decide per each use case; Go for hyper-converged, if benefits outweigh drawbacks.


The next step: Software-Defined DC

Characteristics

Virtualized everything (compute, storage, NW)

Deployable on any HW (components interchangeable)

Unified management

Automation (provisioning, configuration, operation)

Policy-driven (adaptive) resource orchestration

Hyper-converged or converged

Scalable

Optional

Self-service portal

Metering and charging

Simple, fast, flexible, efficient: The right answer to digitalization.

Any IP network

Any storage

Any x86 server

SDDC Platform

Any app


How to Get Your Infrastructure Up and Running?


DIY is error-prone, time-consuming, risky and expensive

DESIGN - Select from myriad of SW and HW products

INTEGRATE - Assemble involving multiple admin domains

TEST - Work off complex test matrix

DEPLOY Integrate into production environment

Do-it-yourself approach Time to production

Pro

ject

sta

rt


Integrated Systems reduce complexity, time, risk and cost

DESIGN - Select from myriad of SW and HW products

INTEGRATE - Assemble involving multiple admin domains

TEST - Work off complex test matrix


Do-it-yourself approach Time to production


Integrated Systems approach

DESIGN - Based on best practices

INTEGRATE - Pre-integrated single order delivery

TEST End-to-end quality assurance

Pro

ject

sta

rt

Solution Development in factory At customer's site


FUJITSU Integrated System PRIMEFLEX

Pre-configured

Pre-integrated

Pre-tested

Reduce complexity, time, risk and cost. Increase operational efficiency. Focus on business.

Ready-to-run and Reference Architectures


FUJITSU Integrated System PRIMEFLEX – Your fast track to DC infrastructures

Characteristics Addressing business-critical use cases

Converged and hyper-converged

Standard sizes ready-to-run and customizable reference architectures

Best practices and experience

Own and partner technologies

Pay as you grow through scaling up and out

Flexible services options

Reduce complexity, time, risk and cost. Increase operational efficiency. Focus on business.


FUJITSU Integrated System PRIMEFLEX: Line-up Focused on what our customers demand for

Your fast track to data center infrastructures.

Virtualization Big Data & Analytics

SAP

Private Cloud HPC


PRIMEFLEX and virtualization

Storage Spaces Direct


VMware Cloud

Foundation

Hyper-converged

Classic converged


PRIMEFLEX and virtualization



VMware Cloud

Foundation

Hyper-converged

Classic converged 2-(3x)64 2-16


Most powerful virtualization platform with excellent price / performance ratio

#1 p

osit

ion

s of

VM

war

e VM

mar

k 2.

x


PRIMEFLEX: SDDC options

Hyper-converged Converged


Summary


Key take-aways

Hyper-converged is a rising star

Many benefits

Springboard to SDDC

But use cases matter

Integrated Systems

Reduce complexity, time, risk, costs

Increase operational efficiency

FUJITSU Integrated System PRIMEFLEX

Converged and hyper-converged

Ideal solution for your use case

Wondering if hyper-converged is your way to go? Have a word with Fujitsu.


Abstract

Hyper-converged infrastructures are the rising stars in data centers.

What is behind this hype?

What are its key benefits and what does it mean in practice?

Is hyper-convergence always preferable a classic IT architecture?

For which use cases is hyper-converged strongly recommended or even a must?

And in which way is hyper-convergence fitting in software-defined data centers?

This session provides appropriate answers.

fujitsu world tour 2017: hyper converged data centre

Technology