Cisco ucs architecture comparison business decision maker

© 2013 Cisco and/or its affiliates. All rights reserved. Other names and brands may be claimed as the property of others.1

Cisco UCS Architecture ComparisonBusiness Decision Maker

Cisco SystemsData Center and VirtualizationUnified Computing System

July 30, 2013


Hide or remove this slide before presenting

This deck is for Cisco Sales and Partner use with all customers.

No NDA required.

This deck is NOT a leave behind.Do not distribute soft or hard copies of

this presentation other than to Cisco Sales and Partners.


THECISCO

UNIFIED COMPUTING SYSTEM DIFFERENCE

CISCO UCS


Legacy Infrastructure Designs

• Infrastructures designed separately – not as a unified system

• Marketed as “converged”, but really management layers on top of multiple infrastructure silos

• Sprawling patchwork of tools, agents and management points

Complexity Drives Up Management Costs

• Rigid models to upgrade and maintain system-level designs

• Multiple tools means multiple points of configuration

• Brittle design with complex inter-dependencies

Legacy Infrastructure and Management

CISCO UCSUNIFIED by DESIGN

Eliminating Silos – Fabric Centric Architecture – Single Point of Mgmt


2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013$0

$50

$100

$150

$200

$250

Power & Cooling Expense

Mgmt. & Admin - Vir-tual Servers

Mgmt. & Admin - Standalone Servers

Server Spending

Cus

tom

er S

pend

ing

($B

)

Data Center EconomicsManagement is the Key Server OpEx driver

Source: IDC, “New Economic Model for the Datacenter”

Billions of Dollars – WW Server Management SpendTotal Server

Related SpendYear Physical Servers Virtual Servers

2003 $ 64 $ 3 $ 135

2007 $74 $ 28 $ 191

2013 $50 $105 $250 31% - 32% 275%- $24 + $ 77 + $ 58

WW Server Related Spend (CapEx + OpEx) - Servers, Power & Cooling, and Mgmt./Administration


Controlling Data Center Cost

Unified Computing System

I/O & Virtualization

Architecture & Scale

Management


Cisco UCS “New” Legacy

Siloed & Complex• Multiple I/O protocols & stranded capacity• High port consumption, no design leverage• Limited & separate physical & virtual port

visibility, minimal control, multiple tools.

Unified Fabric• Single port - LAN, SAN, Mgmt path• Reduced complexity• Physical & virtual port end to end

visibility and control with a single tool

Complex Mgmt Structure• Multiple mgmt tools, multiple interfaces• Every Administrator has multiple tools• Duplicative mgmt points and access,

complicated and inefficient with no scale

Unified Management• Single mgmt tool, single interface• Highly collaborative roles based control• Mgmt interface leveraged across

multiple servers and domains

Scattered, De-centralized Compute• No truly functional identity abstraction• Blade and rack servers segregated, no

identity portability between form factors• Physical & virtual identities independent

Unified Compute• Stateless Computing, abstracted identity• Portable Identities - form factor agnostic,

blade to rack server identity transfer• Physical & virtual functionally combined

Vs.

Management

I/O & Virtualization

Architecture& Scale


The Cisco UCS DifferenceCisco's Unified Data Center strategy unifies physical and virtual infrastructures across data centers.

Delivered more economically without compromising functionality, performance, scalability, operational efficiency or security.

Stateless Computing• Identity = Server Settings and Policies, 127+ parameters & policies• Abstracted Identity = Model-based, GUI driven

Service Profiles PortabilityPortability between blade AND rack servers

Unified Management – Architecture is Key• Centralized Architecture, not de-centralized legacy design

Easy Scaling - Self Aware, Self Integrating, Automated• Form factor agnostic = rack and blade together• Reduced Complexity and Roles Based Access

Servers, LAN, SAN, Management – one tool, one interface.


Blade Architectureand

Scaling


Blade Architecture and ScalingSimpler Design, Scale without Complexity

HP

Growing Capacity Requires Infrastructure Change

Scale Requires Large Increments, 16 blades / 10 RU,Larger Embedded Cost, More

Management Overhead.

Architecture Complex and Cumbersome at Scale

High Top of Rack switch port consumption with increasing

scale.

UCS

Constant Infrastructure With Growth

Scale In Smaller Increments, 8 blades / 6 RU,

Lower Cost, Leveraged Architecture.

User Customizable Architecture. Simple to scale at blade, chassis

and I/O level.

IBM

Growing Capacity Requires Infrastructure Change

Scale Requires Large Increments, 14 blades / 10 RU,

Larger Embedded Cost, Increasing Mgmt Overhead

Architecture Complex and Cumbersome at Scale

High Top of Rack switch port consumption with increasing

scale.

Scaling is a plug and play operation


HP c7000 Platinum Chassisand Virtual Connect


Chassis Mgmt Modules

VC FlexFabric SWITCHES

VCEM License

VCEM License

VCEM License

VCEM License

HP c7000 Platinum Blade ChassisFor UCS Manager parity, you need HP Virtual Connect (VC) Enterprise Manager (VCEM) + HP Insight Control, at the minimum.• Mgmt SW host – Required for SIM & VCEM.• VCEM required on each chassis to move

blade identities (server profiles).• 10 RU chassis. 4 Chassis = 72 slots.

Each Chassis has:• 2 FlexFabric switches per chassis• 2 x Mgmt Modules per chassis.• = 4 mgmt points.

4 Mgmt Ports in EVERY chassis – minimum.

Mgmt SW host required

Chassis Mgmt Modules

1Gb Enet Mgmt

ToR switches are needed to connect multiple chassis. Switches are redundant

10Gb Enet

FC Switch

VC FlexFabric SWITCHES

VCEM License

VCEM License

VCEM License

VCEM License

BackFront


HP c7000 Platinum Chassis5 Gbps Enet / blade (+ 2 Gbps FC / blade)

10Gb LAN

FC SAN

Mgmt SW host

1 chassis – 16 blades

16 BL460c Gen8 blades in one chassis.• FlexibleLOM – FlexFabric 10Gb dual port

2 x HP FlexFabric switches, Each switch has a maximum 8 uplink ports4 – 10 Gb Enet4 – 10 Gb Enet or FC4/8

Illustrated here: 80 Gbps Enet total (4 – 10Gb ports / switch x 2 switches)

÷ 16 blades in each chassis 5 Gbps Enet only / blade leaving the chassis

32 Gbps FC (2 FC8 ports / switch x 2 switches)

÷ 16 blades 2 Gbps FC only / blade leaving chassis

5 Gbps of Enet only / blade – dedicated Enet2 Gbps FC only / blade - dedicated FC7 Gbps Total I/O per blade leaving chassis

Mgmt switch

10Gb

FC8

1Gb

FC8

1Gb

10Gb

HP c7000 Platinum chassis• 10 fans, 6 power supplies• 16 Insight Control Licenses• 2 chassis management modules• 2 FlexFabric switches + VCEM• 16 - HP BLc460 Gen8 blades

Chassis Cable Count 8 - 10Gb Enet

4 - FC8

2 - chassis mgmt

14 cables / chassis

16 blades / chassis

1/10Gb FC Mgmt


HP c7000 Platinum Chassis5 Gbps Enet / blade (+ 2 Gbps FC / blade)


******************This is a lot of cables for only 7 Gbps of I/O / blade

• 2 Gbps of FC / blade• 5 Gbps of Enet / blade

4 chassisx 14 Cables each 56 cables 56 ToR switch ports$ $ $ $ $

5 Gbps Enet / blade not enough I/O?

4 chassis – 64 blades16 management points – 4 per chassis.

We aren’t managing the blades yet.

10Gb LAN

FC SAN

Mgmt SW host

Mgmt switch

10Gb

FC8

1Gb

FC8

10Gb

1Gb


IBM Flex System Chassisand Flex System Manager


Chassis Mgmt

Modules

CN4093 10Gb SWITCHES

IBM Flex System Blade ChassisFor UCS Manager parity, you need IBM Flex System Manager (FSM) at the minimum.• FSM Mgmt Node – Required for every 4

chassis.• FSM Mgmt Node – NOT REDUNDANT.• FSM license required for every chassis.• 10 RU chassis. 4 Chassis = 56 slots. Only 55

Compute

Each Chassis has:• 2 CN4093 switches per chassis• 2 x Mgmt Modules per chassis.• = 4 mgmt points

4 Mgmt Ports in EVERY chassis – minimum

FSM License

13 compute slots &

1 mgmt node

Chassis Mgmt

Modules

CN4093 10Gb SWITCHES

BackFront

FSM License

FSM License

FSM License

1Gb Enet Mgmt

Switches are needed to connect multiple chassis.Switches are redundant

10Gb Enet

FC Switch

14 compute slots


IBM Flex System Chassis5.7 Gbps Enet / blade (+ 2.3 Gbps FC / blade)

10Gb LAN

FC SAN

Mgmt switch

IBM Flex System chassis• All fans• All power supplies• 2 chassis management modules• 2 CN4093 switches• 1 IBM FSM Mgmt Node: chassis 1 only• 14 - x240 (except chassis #1 with 13 + FSM)

10Gb

FC8 FC8

10Gb

1Gb

Chassis #1 – 13 blades (compute)1 FSM Mgmt Node (1 required per 4 chassis)13 – Flex System x240 compute nodes, each with

• 1 – CN4054 10Gb quad port (FCoE upgrade)2 x CN4093 switches, each base switch comes with:

• 14 downlinks (1 per blade) and• 2 x 10Gb uplinks• 6 Omniports uplinks, ports can be 10Gb or FC4/8• 4 Omniports used here

Illustrated here: 80 Gbps Enet total (4 10Gb ports / switch x 2 switches)

÷ 14 blades (14 in compute chassis)

5.7 Gbps Enet only / blade leaving the chassis

32 Gbps FC total (2 FC8 ports / switch x 2 switches)

÷ 14 blades 2.3 Gbps FC only / blade leaving chassis

5.7 Gbps of Enet only / blade – dedicated Enet2.3 Gbps FC only / blade – dedicated FC8 Gbps Total I/O per blade leaving chassis

Chassis Cable Count 8 - 10Gb Enet

4 - FC8

2 - chassis mgmt

14 cables / chassis

14 blades / chassis

1/10Gb FC Mgmt



10Gb LAN

FC SAN

Mgmt switch

1Gb

FC8FC8

10Gb

10Gb

5.7 Gbps of Enet only / blade2.3 Gbps FC only / blade 8 Gbps Total I/O per blade leaving chassis

You can add 2 more 10 Gb Enet connections per switch, 40 Gbps per chassis

80 (original Enet capacity)+ 40 (new 2 x 10 Gb per switch “ ”) 120 Gbps Enet leaving chassis÷ 14 blades in each chassis 8.6 Gbps / blade

The native ports on the CN4093 switches are maxed out.

If you need I/O, more uplinks, there are two options:

1. Buy upgrades for both switches – retail at $10,999 each = $21,998

2. Buy another pair of switches – retail at $20,899 each = $41,798. This option requires more mezz cards as well - $1,868 x 14 blades = $26,152; TOTAL to add switches is $67,950

Chassis Cable Count12 - 10Gb Enet

4 - FC8

2 - chassis mgmt

18 cables / chassis

14 blades / chassis

IBM pricing publically available on 07/04/2013.



10Gb LAN

FC SAN

4 chassis - 55 blades [13 + (3 x 14)], compute(IBM Flex System Manager domain maximum)

******************This is a lot of cables for 10.8 Gbps of I/O / blade

• 2.3 Gbps of FC / blade• 7.5 Gbps of Enet / blade

This ends up being:2 x mgmt cables4 x FC8 cables12 x 10Gb Enet cables18 Cables for each chassis1.28 cables per blade server

4 chassis72 cables 72 ToR switch ports 8 management switch ports$ $ $ $ $

1Gb

Mgmt switch

10Gb10Gb

FC8FC8


Cisco UCS Chassis


Cisco UCS Blade Chassis• No Extra Mgmt SW / Hardware needed.• No “per chassis” licensing needed or required.• UCS Management is FULLY REDUNDANT.• 1 to 20 chassis or 160 RACK or BLADE servers.• 2 x UCS Fabric Interconnects (FI) required.

48 or 96 port models – 10 Gbps FCoE.•All Mgmt SW (UCS Manager) is included in FIs.

• UCS Fabric Interconnects are Active / Active Cluster= 1 mgmt point for ALL chassis & rack servers.

• Each UCS 2208 has 8 x 10Gbps FCoE ports (management path included).

• UCS 2204 version has 4 ports each.• UCS 2208 / 2204 are Line Cards are NOT

switches. They are remote line cards for the Fabric Interconnects and are not a mgmt point.

BackFront

UCS 2208 – LINE CARD10 Gbps I/O & Mgmt

Up to 20 blade chassis (160 blade servers) –

Mix Blade AND Rack Servers – up to 160 servers total.

All in

One Mgmt Tool, One Mgmt Interface

One Mgmt Domain

UCS FI Fabric Interconnects Required

8 compute slots

Up to 20 Chassis

Up to 20 Chassis.

Up to 160 Blade or Rack Servers


Cisco UCS Cisco UCS chassis (qty 2)• All fans• All power supplies• 2 – UCS 2208 I/O modules per chassis• 16 – B200 M3, 8 per chassis.


16 B200 M3 blades, 8 per chassis.• mLOM UCS 1240 VIC – 4 x 10Gb FCoE ports

UCS 5108 chassis, each with 2 x 2208 I/O modulesEach 2208 has 8 x 10Gb FCoE ports = 80 Gb each

Illustrated here:40 Gb (2 x 10 Gb ports per module)÷ 8 blades 5 Gb / blade leaving chassis

2 x UCS 6296UP Fabric Interconnects (FI)96 Universal Ports each for I/OUniversal ports for 10 Gb / FCoE / FC4/8Use for Southbound (to chassis) or Northbound

5 Gbps / blade5 Gbps FCoE per blade leaving chassis

All I/O is available to all blades in the chassis20 Gb minimum available from each bladeFC is prioritizedQoS is set per blade by admins to meet needs

10Gb LAN

FC SAN

4 x 10Gb4 x FC8 4 x FC8

4 x 10Gb

FCoEFCoE

UCS 6296 FI

1/10Gb, FCoE 1/10Gb FC


Cisco UCS10Gb LAN

FC SAN

10Gb FC8 FC810Gb

FCoE

8 chassis - 64 BladesLess than ½ of the UCS Manager Domain limit:

Some chassis / some blades, may need more I/O than others.

Add I/O from the chassis to the FI = Add cables “ ”

Get up to 80Gbps per blade – Your choice

Add more Northbound I/O from the FI= Set the port characteristics, add cables

Cisco UCS has:• No requirement for blades to be identically

configured.• No need to add costly “intra-chassis” switches just

to have or add more I/O on a few blades.• No requirement for chassis to be identically

configured.

UCS 6296 FI

FCoE

Uplink type and count – Variable by YOUR I/O Requirements

Uplink type and count – Variable by YOUR I/O Requirements


Cisco UCS10Gb LAN

FC SAN

10Gb FC8 FC810Gb

FCoE

UCS 6296 FI

FCoE

Uplink type and count – Variable by YOUR I/O requirements.

Adding chassis is easy:Connect I/O cables to Fabric Interconnects20 Chassis – 160 Blade servers No interdependencies


I/O and Virtualization


I/O and VirtualizationUnification Reduces Complexity

HP

Limited visibility of virtual server I/O. Added software required.

Only partial I/O identity with deployment. Deploying servers

very manual and time consuming.

Growing capacity increases complexity

Scale requires large hardware increments including high ToR

switch port consumption.

UCS

Full Port to Port visibility for both physical and virtual servers. No

added cost

UCS Automated Deployment / Provisioning includes I/O

mapping, policies and security.

Unification yields constant, leveraged infrastructure.

Scale in smaller increments, leveraging existing infrastructure. Plug and Play to increase chassis

and blade I/O.

IBM

Limited visibility of virtual server I/O. Added software required with

additional cost.

Only partial I/O identity with deployment. Deploying servers

very manual and time consuming.

Growing capacity increases complexity

Scale requires large hardware increments including high ToR

switch port consumption.


HP VC FlexFabric 10Gb/24- Port Module

X 1 X 2 X 3 X 4 X 6X 5 X 7 X 8

HP I/O Sacrifice for FlexFabric Features No Separate Management Ports

4 x 10 GbEports only

2 x 10 GbEor FC ports

2 x 10 GbEor FC ports

HP prices publically available 07/04/2013

FlexFabric mezz card = $849 each

16 servers/chassis x 16

16 mezzanine cards(base warranty/support)

= $13,584

10 Gb — 8 portsmax. per switch

= 80 Gb/s

x 2 switches(8 x 10 Gb/s port — max I/O)

= 160 Gb/s

Pair of switches(base warranty/support)

= $36,998

$50,842 to add additional I/O

Server I/O CostsChassis I/O Costs ( per chassis) – NO SFP+ included

HP VC FlexFabric10Gb/24- Port Module


1 2 3 7

11 12

13

14

15

16

17

18

19

20

21

229

10

56

34

783

45

6

78

91

0

IBM I/O Sacrifice for Flex System Features

Lots of License and Upgrade Costs

IBM prices publically available 07/04/ 2013.

Flex System CN4054 10Gb Virtual Fabric Adapter

$1,099 each

CN4054 Virt Fab Adapter, SW upgrade to provide FCoE & iSCSI

+ $ 769 each

Total per mezz card per server = $1,868

14 servers/chassis x 14

14 mezzanine cards(base warranty/support)

= $26,152

Native: 10 Gb — 8 ports = 80 Gb/s

x 2 switches(8 x 10 Gb/s port — max I/O)

= 160 Gb/s

Pair of switches(base warranty/support)

= $41,798

2 x 10 GbESFP+ portsStandard

2 x 40 GbEQSFP+ ports

Requires Upgrade #1

12 Omni ports SFP+ (10Gb or FC 4/8)6 Standard,

6 Require Upgrade #2

$67,950 to for additional I/O

Mgmt portStandard

Server I/O CostsChassis I/O Costs ( per chassis) – NO SFP+ included

IBM Flex SystemCN4093


Cisco VIC is really like a “Flex-256” adapter that includes multiple vHBA support

Cisco VIC vs. HP FlexFabric Adapter

OS

CiscoBladeServer

Single lane of 10Gb/s Ethernet

for each Port

VIC mLOM or Mezz. Adapter

vNIC11

vNIC12

vNIC13

vNIC5

vNIC6

vNIC7

vNIC8

vHBA9

vHBA10

vNIC253

vNIC254

vNIC255

vNIC256

vHBA1

vHBA2

vHBA3

vHBA4

vNIC20

vNIC21

vNIC22

vNIC14

vNIC15

vNIC16

vNIC17

vNIC29

vNIC30

vNIC31

vNIC23

vNIC24

vNIC25

vNIC26

vNIC38

vNIC39

vNIC40

vNIC32

vNIC33

vNIC34

vNIC35

vNIC47

vNIC48

vNIC49

vNIC41

vNIC42

vNIC43

vNIC44

vNIC56

vNIC57

vNIC58

vNIC59

vNIC50

vNIC51

vNIC52

vNIC53

Physical CNA Port 1

Physical CNA Port 2

Physical CNA Port 1

Physical CNA Port 2

vNIC18

vNIC19

vNIC27

vNIC28

vNIC36

vNIC37

vNIC45

vNIC46

vNIC54

vNIC55

FEX A

Fabric interconnect A

FEX B

Fabric Interconnect B

OS

Flex NIC1

HPBladeServer

Single lane of 10Gb/s Ethernet

for each Port

Flex HBA2

Flex NIC3

Flex NIC4

Flex NIC5

Flex HBA6

Flex NIC7

Flex NIC8

FlexFabric LOM or Mezz. Card

VC FlexFabric Module(Bay 2)

Physical NIC Port 1 Physical NIC Port 2

VC FlexFabric Module(Bay 1)

http://blog.tmcnet.com/blog/tom-keating/images/cisco-logo.gif

http://images.google.com/imgres?imgurl=http://www.engineering.cornell.edu/diversity/office-diversity-programs/summer-programs/highschool-programs/catalyst/images/hp_logo_1.jpg&imgrefurl=http://www.engineering.cornell.edu/diversity/office-diversity-programs/summer-programs/highschool-programs/catalyst/index.cfm&usg=__BCEJcWQJWS79gmyIXo1PBXIbgQM=&h=569&w=900&sz=51&hl=en&start=2&tbnid=SRQw8He8D6YA_M:&tbnh=92&tbnw=146&prev=/images?q=hp+logo&gbv=2&hl=en


LatencyUp to 40% lowerend-to-end latency

ThroughputUp to 10% moreAt 30% lower CPU utilization

Application Performance

Up to 15% more (Database workload)

VM-FEXHighest Performance Virtual Networking

Software Switch

vNIC

vNIC

VM-FEX(Hypervisor Bypass)

vSphere 5

vEth

vEth

Cisco UCS Delivers Enhanced Performance

Others

vSphere 5

vSwitch


Cisco UCS™ 6200 and 2200 with Unified Ports

Typical Deployments

UCS-FI-6248UP

48 Port Fabric Interconnect

• Performance for typical deployments,• 1TB throughput,• 48 ports in 1RU,• Infrastructure agility with Unified Ports.

High End Deployments

UCS-FI-6296UP

96 Port Fabric Interconnect

• High Application performance ,• 2TB through put,• High workload density 96 ports in 2RU,• Infrastructure agility with Unified Ports.

16 Port I/O Module

UCS-FI-2204XP

• 80G/ chassis, • 20Gb to the Blade each, 40Gb total per

blade,• Improved Utilization with Port Chnnels.

32 Port I/O Module

UCS-IOM-2208XP

• 160G/ chassis, • 40Gb to the Blade each, 80Gb total per

blade, for burst traffic,• Improved Resiliency ,• Improved Utilization with Port Channels.

UCS FabricInterconnects

UCS FEX I/O Modules

Cisco UCS Fabric Infrastructure Portfolio


Blade Management


Blade ManagementUCS - Less Complexity, More Flexibility, Easy Scale

HP

Back of each blade chassis has a “rack’s worth of infrastructure”

Blade and Rack servers require separate management

Back of each chassis is a hardware profit center

Adding chassis adds a “rack’s worth of infrastructure” burden

UCS

One infrastructure for multiple blade chassis and racks

One Management interface for multiple blade chassis AND rack

servers

Low cost FEX integrates Management and I/O (Enet, FC

and Mgmt)

IBM

Back of each blade chassis has a “rack’s worth of infrastructure”

Blade and Rack servers require separate management

Architecture is a Software Profit Center. Back of each chassis is a

hardware profit center

Adding chassis adds management software burden

and a “rack’s worth of infrastructure” burden

127+ Server ID Settings—Completely Automated Including

Firmware and I/O Devices


Cisco Service Profiles: Heart of Unified Model-Based Management

• Allows YOU to define the “to-be” server, NOT settle for the “as is” server

• Created through Cisco UCS Manager

• Configure once then reuse

• Templates as Best practices

• Blade and Rack Servers – Service Profiles are Form Factor Agnostic

NIC MACsHBA WWNsServer UUIDVLAN AssignmentsVLAN TaggingFC Fabrics AssignmentsFC Boot ParametersNumber of vNICsBoot orderPXE settingsIPMI SettingsNumber of vHBAsQoSCall Home

Template AssociationOrg & Sub Org Assoc.Server Pool AssociationStatistic ThresholdsBIOS scrub actionsDisk scrub actionsBIOS firmwareAdapter firmwareBMC firmwareRAID settingsAdvanced NIC settingsSerial over LAN settingsBIOS SettingsMore….

CISCO UCS SERVICE PROFILE

LAN

SAN


Integrated Partner SolutionsBroad Tool Choice, Powerful Integration with UCS

HP Server Platforms

UCS CLI

UCS GUIUCS

Blade Servers

No Direct HP Hardware

Management No API

No Direct HP Hardware

Management No API

Direct Hardware Management via XML / API

Direct Hardware Management via XML / API

UCS Rack Servers

Cisco Unified Computing System

UCS Director


Simpler ArchitectureHP doubling servers = doubling touches; UCS = 1 touch point

64 Blades – 4 x HP c7000

Fabric Interconnects 0

Intra Chassis Switches 8

Chassis Mgmt Module 8

Total Mgmt Points 16

80 Blades – 10 x Cisco UCS 5108

Fabric Interconnects 2

Intra Chassis Switches 0

Chassis Mgmt Module 0

Total Mgmt Points 1

Mgmt switchLAN

LANSAN A

SAN B

Mgmt switch

Mgmt switch

LAN

SAN A

SAN B

LANMgmt switch


16 blade servers0 rack servers

Virtual Connect Enterprise ManagerVC Enterprise Manager $$

System Insight Manager(SIM)

HP Insight Control $$

UCS—More Flexible, Less Complexity

HP Server Hardware Management

Multiple Layers of Software Required

HP c7000


UCS Manager 1 Console

No Added CostRack and Blade Together

Cisco UCS

HP iLO Advanced for BladeSystem

Virtual Connect Manager

Onboard Administrator













Separate Management - Every Chassis, All SoftwareSeparate Enet & Fibre Channel I/O leaving the chassis


Up to 160 serversBlade or Rack mount

Unified Compute Unified Management Unified Networking


Virtual Connect Enterprise Manager $$$$

HP Insight Control $$$$

System Insight Manager(SIM)

Virtual Connect Enterprise Manager $$$$$

HP Insight Control $$$$$

HP System Insight Manager(SIM)

24 Blade Servers6 Rack Servers


Separate Management

Stack

The Cisco UCS Management Difference

Blade Chassis Management

Rack ServerMgmt

Enet Switch Mgmt FC Switch Mgmt

“New” Legacy Servers

Switch Mgmt

Blade Server Mgmt -

16 Blades

1/10Gb, FCoE 1/10Gb FC

Cisco Unified Computing System

Cisco UCS provides a single management tool (UCS Manager)• Unified Compute – Abstracted Server Identities to Service Profiles 127+ identity settings

Form Factor agnostic – blade or rack – with portability back and forth• Unified I/O – Server, LAN, SAN and Management into one interface• Unified Management – unified across a distributed environment

Cisco UCS Manager – 160 servers

Fabric Interconnects

Blade Chassis

Rack Servers

Fabric Extenders

Increments of scale – leveraged architecture, not

duplication.

Increment of scale – everything duplicated:

management, switches, etc.

Blade Chassis


Cisco UCS Central

Increasing ScaleUCS has 160 server increments, not 16 blades

“New” Legacy Servers

Separate Management

Stack



Rack ServerMgmt

Enet Switch Mgmt FC Switch Mgmt

Switch Mgmt

Blade Server Mgmt -

16 Blades

Switch Mgmt

Blade Server Mgmt -

16 Blades


Switch Mgmt

Blade Server Mgmt -

16 Blades

Cisco UCS Manager – 160 serversFabric Interconnects

Blade Chassis& Servers

Rack Servers

Fabric Extenders

Blade Chassis

Cisco UCS Manager – 160 serversFabric Interconnects

Blade Chassis& Servers

Rack Servers

Fabric Extenders

Blade Chassis

Multiple UCS

Manager Domains

1/10Gb, FCoE

1/10Gb

FC


XML API

• Blade and Rack Servers in the same domain – Form Factor Agnostic• Standards-based XML API presents bidirectional single interface to entire solution• UCS offers the customers the broadest choice of Cisco or 3rd party management tools

10,000 UCS SERVERS - MONITOR and MANAGE - SEAMLESSLY

UCS CLI + UCS Management + UCS Director

goUCS Automation Tools

CDN .NET/Windows PowerShell

Python, Perl XML

Customer

UCSMData Center 1

UCSMSingle UCS Domain

UCS Is Redefining Server Management

UCS ManagerCLI

UCSMData Centers 3, 4 ....

UCSMData Center 2

Third Party

UCSCentral

UCSDirector(Cloupia)


Total Cost of Ownership


TCO

HP

Costly to add more chassis and I/O

HP “accidental mini-rack” chassis design has high cost

burden to scale

Through-put trade off for features

HP just announced a new chassis with no upgrade for

older chassis.

UCS

Efficient and Effective, low cost I/O additions

UCS delivers lower TCO by design

with easy, lower cost scaling

No sacrifice of function for features

UCS chassis has the future built in today

IBM

Costly to add more chassis and I/O

IBM Flex System is more of the same with high cost burden to

scale

Lots of cost adders for limited additional functionality.

IBM new Flex System chassis is a software selling mechanism.


UCS & HP: Infrastructure Scaling Cost

16

$63,1242 x UCS 6248UP FI

2 x UCS 5108 chassis

HP c7000 Platinum chassis, each with:• 10 fans, 6 power supplies & cords• 16 Insight Control Licenses• 2 Enclosure Management Modules• 2 FlexFabric switches • HP VC Enterprise Manager

32

48

64

UCS 6248UP Fabric Interconnects , each with:• All fans, power supplies & cords, and acces kits

Cisco UCS chassis, each with:• 8 fans, 4 power supplies & cords• 2 – UCS 2208 I/O modules per chassis• 4 – 10Gb SFP+ cables

$63,124

$60,545HP c7000 chassis




$60,545

$121,090

$181,635

$86,203

$ 109,282

$23,0792 x UCS 5108 chassis



16 server capacity

16 server capacity

$ 3,784 / server $ 2,068 / server

Cisco UCS45% less than HP

64Servers

$242,180 $132,362$242,180 $132,362

All pricing publically available on 07/04/2013.


UCS & IBM: Infrastructure Scaling Cost

16 $63,1242 x UCS 6248UP FI

2 x UCS 5108 chassis

UCS9 more servers$142,000 less

$274,877 $132,362

32

48

64

UCS 6248UP Fabric Interconnects, each with:• All fans, power supplies & cords, and acces kits

Cisco UCS chassis, each with:• 8 fans, 4 power supplies & cords• 2 – UCS 2208 I/O modules per chassis• 4 – 10Gb SFP+ cables

$63,124

$78,656IBM Flex System chassis13 servers (14 – 1 FSM node)All other chassis = 14 slots

$65,407IBM Flex System chassis14 slots

$65,407IBM Flex System chassis

$65,407IBM Flex System chassis

$78,656

$144,083

$209,470

$86,203

$ 109,282



$23,079UCS 5108 chassis

13 server capacity

16 server capacity

$ 4,998 / server$ 2,068 / server

Cisco UCS52% less than IBM

13

27

41

55

IBM Flex System chassis, each with:• All fans, power supplies & cords• 2 – chassis management modules• 2 – CN4093 10Gb switches• 1 – Flex System Manager license• 1 – IBM FSM Mgmt Node –

Chassis 1 only

14 server capacity

6455

$274,877 $132,362

All pricing publically available on 07/04/2013.


UCS = Better, Easier, Simpler Architecture

No Infrastructure Penalty to Scale

Cisco pricing MSRP on 07/04/2013.HP pricing publically available on 07/04/2013. IBM pricing publically available 07/04/2013.All pricing is for blade chassis and networking only. Servers are not included.

13 14 16 17 24 32 40 48 55 56 64 72 80 88 96$0

$100,000

$200,000

$300,000

$400,000

$500,000

$63,641 $75,180$98,259

$127,256

$144,713$179,627

$214,541

$60,545

$121,090$181,635

$242,180

$302,725

$363,270

$78,656

$144,063

$209,470

$274,877

$353,533

$418,940

$484,347

Total Number of Chassis Blade Server Slots

Cisco UCS

HP c7000

IBM Flex

Cha

ssis

and

I/O

Cos

t

BLADE CHASSIS SAVINGS AT SCALE—BLADE SLOT SOLUTIONUCS: UCS 5108 chassis with UCS 6248 FI (two uplinks per FEX)

HP: HP c7000 Plat chassis w/ 2x VC Flex Fabric and 16x HP IC. Price includes HP VCEM each chassis

IBM: IBM Flex Chassis with 2x CN4093 switches, one Mgmt Node every 4 chassis, FSM license each chassis

HP is $45,910 more to get ready to add a 17th server.

Cisco UCS is 38% less than HP.

IBM Flex is $80,422 more to get ready to add a 14th server.

Cisco UCS is 56% less than IBM.


No Compromise – full chassis adds Cisco Solution TCO advantage increases at scale

$0

$200,000

$400,000

$600,000

$800,000

$1,000,000

$1,200,000

3 Year Warranty

3 Year Power Costs

Server Deployment

Cabling

TOR Switches

Chassis, Interconnects, Hdw Mgt

Server Hardware

8 16 32 48 64 80# of Servers

HP Trend Line

Cisco Trend Line

ServersHP: BL460 Gen8Cisco: B200 M3

All chassis fully populated with servers (starting at qty 16)

Each server has two E5-2620 Intel Xeon processors with 64GB memory (eight 8GB DIMMs)

CiscoHP CiscoHP CiscoHP CiscoHP CiscoHPCiscoHP

64GB

HP retail and Cisco MSRP pricing on 4/3/2013


Performance


Higher Performance

HP

Updating requires multiple touches

Indeterminate latency

60 Gbps / blade, expensive with more hardware to manage

16 DIMMs in the BL460 Gen8

UCS

Cisco single touch updates / deploys faster

Identical latency between blade servers

80 Gbps per blade, for much less. Leverages existing infrastructure.

24 DIMMs in the B200 M3

IBM

Updating requires multiple touches

Indeterminate latency

80 Gbps / blade, expensive with more hardware to manage

24 DIMMs in x240 but throughput requires costly switch upgrades


Cisco UCS Performance-73 RecordsA History of World Record Performance on Industry Standard Benchmarks

Cisco UCS Benchmarks that held world record performance records as of date of publication

Best HPC Performance

SPECompLbase20012-socket B200 M2

SPECompMbase20012-socket B230 M2


SPECompMbase20014-socket C460 M2

SPECompMbase2001 2-socket B200 M2


LinPack2-socket B200 M2

LS-Dyna4-socket C460 M1


SPECompMbase20012-socket B200 M2




SPECompLbase20012-socket C220 M3

Best Database Performance

TPC-COracle DB 11g & OEL

C250 M2

TPC-H 1000GBMicrosoft SQL Server

C460 M2

TPC-H 100GBVectorWiseC250 M2

TPC-H 300GBVectorWise

C250 M2

TPC-COracle 11g C240 M3

Best Enterprise Application Performance

Oracle E-Business Suite Medium Model Payroll

Batch B200 M2

Oracle E-Business Suite Xtra Large Model Payroll

B200 M3

Oracle E-Business Suite Medium Model Payroll

Batch B200 M2

Oracle E-Business Suite Medium Model

Order-to-Cash B200 M2

Oracle E-Business Suite Large Model

Order-to-Cash B200 M3

Oracle E-Business Suite Ex-large Model Payroll

Batch B200 M2

SPECjEnterprise2010 Overall B440 M1

SPECjEnteprise20102-node B440 M2

Oracle E-Business Suite Xtra Large Model Payroll

Batch B230 M2

Oracle E-Business Suite XL Model Payroll B200

M3

Oracle E-Business Suite Large Model Order-To-

Cash B200 M3

Oracle E-Business Suite XL Model Payroll B200

M3

Oracle E-Business Suite Large Model Order-To-

Cash B200 M3

Best CPU Performance SPECfp_rate_base2006

2-socketC260 M2

SPECint_rate_base2006 2-socketC260 M2

SPECint_rate2006X86 4-socket

C460 M2

SPECint_rate_base2006 X86 4-socket C460 M1

SPECint_rate_base2006 X86 2-socket

B200 M2

SPECint_rate_base2006 X86 2-socket B200 M1

SPECfp_rate_base2006 X86 2-socket B200 M2

SPECint_rate_base2006 X86 2-socket

B200 M2

SPECfp_rate_base2006 X86 4-socket C460 M1

SPECfp_rate_base2006 X86 2-socket B200 M1

SPECfp_rate_base2006 X86 2-socket C220 M3

SPECint_rate_base2006 X86 2-socket C220 M3

SPECfp_base2006X86 2-socket

C220 M3

SPECint_rate_base2006X86 2-socket

C220 M3

Best Enterprise Middleware Performance

SPECjbb20052-socket C260 M2

SPECjbb20052-socket B230 M2



SPECjbb2005X86 2-socket B200 M2

SPECjbb2005X86 4-socket C460 M1

SPECjAppServer2004 2-node B230 M1



SPECjAppServer20041-node 2-socket C250 M2



Best Virtualization & Cloud Performance

VMmark 2.0Overall B200 M2

VMmark 2.12-socket Blade B200 M2

VMmark 1.x2 –socket Blade B230 M1

VMmark 1.x Overall C460 M1

VMmark 1.x 2-socket B200 M1


VMmark 1.xOverall C460 M1

VMmark 1.xBlade Server B440 M1


VMmark 2.1Overall C460 M2

VMmark 2.1 Two–node 4-socket C460

M2

VMmark 2.14-socket C460 M2

VMmark 2.1 Two–node 2-socket B200

M3

VMmark 2.1 Eight–node 2-socket

B200 M3

VMmark 2.5 Two-node 2-socket C240

M3


High PerformanceSingle Latency with UCS

HP and IBM Talk About East to West I/O Traffic. They say that:• 80% of all blade I/O traffic is East/West within a single chassis, not North/South (N/S); North/South is I/O traffic

leaving any chassis

• Only 20% of all I/O traffic actually leaves the chassis

Is This Realistic? What Does This Mean for Actual Users?• All the blades that “need to talk to each other” (cross talk) need to be in the same chassis;

Is this realistic?

• 20% North/South I/O traffic seems very low for typical data center production environments. Is this realistic?

• Web servers, file/print servers, DB servers, etc., all generate significant N/S traffic to LAN and SAN

• Virtualization in data centers today means that the mix of VMs on any physical server mitigates against “cross talk” servers all being in the same chassis.

• What happens to latency dependent application performance when you migrate a server identity to a different blade in a different chassis? Or VC domain?

• Does the latency change impact your required performance?

• How does this affect the usefulness of blade identity portability in HP solutions?

High Performance Needs Low, Defined and Dependable Latency


Faster, More FlexibleUCS Automated Deployment

20 20

27

38

0

10

20

30

40

50

1-blade scenario 2-blade scenario

Min

ute

s

Cisco UCS solution

HP solution

The Cisco UCS Solution Reduces Time

10 14

24

42

0

10

20

30

40

50

1-blade scenario 2-blade scenario

Ste

ps

Cisco UCS solution

HP solution

The Cisco UCS Solution Reduces Complexity

Cisco UCS - Model-based management speeds deployment Fewer touch points reduces errors

http://www.youtube.com/watch?v=nijWlNzSgCQ

http://www.youtube.com/watch?v=nijWlNzSgCQ


Existing Maintenance

Budget100%

Cisco Unified Computing SystemChanging the Economics of the Data Center

40–50%Maintenance Now

Funded Project

Funded Project

Funded Project

NEW IT Projects – No Additional Budget

TCO/ROI Advisor: http://www.cisco.com/en/US/solutions/ns340/ns517/ns224/tools/data_center_value_zone.html#~Overview

http://www.cisco.com/en/US/solutions/ns340/ns517/ns224/tools/data_center_value_zone.html


Blade Server Marketplace


Customers Have Spoken

Maintained #2 in N. America (27.9%) and #2 in the US (28.3%)1

UCS x86 Blade servers revenue grew 35% Y/Y in Q1CY131

Advanced to #2 worldwide in x86 Blades with 19.3%

UCS momentum is fueled by game-changing innovation; Cisco is quickly passing established players

UCS #2 in Only Four Years

X86

Ser

ver

Bla

de

Mar

ket

Sh

are,

Q1C

Y13

1

UCS #2 with 26.9%

Source: 1 IDC Worldwide Quarterly Server Tracker, Q1 2013, May 2013, Revenue Share

HP

Cisco

IBM

Dell

NEC

Hitachi

Fujitsu

Oracle

0 5 10 15 20 25 30 35 40 45 50

Worldwide

UCS #2 19.3%

Oracle

SGI

Dell

IBM

Cisco

HP

0 5 10 15 20 25 30 35 40 45


Cisco is a Leader in the 2013 Gartner Magic Quadrant for Blade Servers

Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings. Gartner research publications consist of the opinions of Gartner's research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose.

This graphic was published by Gartner, Inc. as part of a larger research document and should be evaluated in the context of the entire document. The Gartner document is available upon request from Gartner 2013 Magic Quadrant for Blade Servers

Read the Full Report here:

Gartner 2013 Magic Quadrant for Blade Servers

By Andrew Butler and George J. Weiss, G00250031, April 29, 2013, © 2013 Gartner Inc

http://www.gartner.com/reprints/cisco-v3-us-7?id=1-1FGKRJ6&ct=130506&st=sb





Worldwide Market Share of WW x86 Blade

Total Factory RevenueMarket Share of WW x86

Blade Total Units

Revenue Share Change Unit Share Change

Cisco + 5.3% + 6.0%Dell - 0.2% + 0.6%HP - 1.5% + 1.1%IBM - 2.2% - 6.0%

All Others - 1.4% - 1.7%

Market Share Changes – Q1’12 to Q1’13Customers are voting for UCS

USA Market Share of USA x86 Blade

Total Factory RevenueMarket Share of USA x86 Blade Total Units

Revenue Share Change Unit Share Change

Cisco + 5.6% + 8.6%Dell - 1.2% + 1.7%HP - 0.3% + 2.8%IBM - 3.0% - 12.0%

All Others - 1.1% - 1.0%

X86 Blade Market Share Numbers – WW and USQ1 2012 to Q1 2013 Share Changes

Source: IDC Worldwide Quarterly Server Tracker, Q1 2013, May 2013,


Thank you.Thank you.

Technology

Cisco ucs architecture comparison business decision maker