Global Networks and Data Centers Briefing...Catalyst 6509 = 112 10G ports Nexus 7018 = 512 10G ports...

May 27, 2011

Global Networks and Data Centers Briefing

Target State T ( ~ FY 13 )

Netherlands – Metro-based DC Pair (Tier-III)

Single-Instance Order Management (OM/AR)

Cisco International

AsiaPAC TBD – Single DC (Tier-III) + land

Continental hub for SaaS, Unified Communications and software development

1 A 1 x Type - A ( Tier - III )

2 A sc 2 x Type - A at Synchronous Capable Distance

1 A

2 A sc

2 A sc

1 A

E

E Early Adopter DC

(~Uptime Tier-II) (~Uptime Tier-III)

B

B Type - B ( Tier - II ) 1 x Type - A ( Tier - III )

B

2 x Type - A at Synchronous Capable Distance

B

40 ms rtt

B B

B

B

B

B

B B

Distributed standalone DCs (Tier-II)

Latency-sensitive software development at lower availability

Richardson & Allen (TX) – Metro-based DC Pair (Tier-III) + Carrollton Co-Lo

Global hub for business applications Continental hub for SaaS and communications

Global Disaster Recovery Strategy

Short to Mid-Term: Leverage Current Assets

Long-Term: DR + SaaS growth

1 x Type - A ( Tier - III )


BC / DR Plan Type - B ( Tier - II ) 1 x Type - A ( Tier - III )


BC / DR Plan

1 A

Mountain View, CA – Early Adopter DC

Production New Product/Service Introduction

Technology Leadership Business Strategy Cisco Green Initiative

Acceleration

Automation

Virtualization

Next-Gen DC

Networking

Adoption

Curve

Management

New

Opportunities

Accelerated Adoption

• Shared (leveraged) Infrastructure

• Aligned Architecture – application, business, infrastructure

• Reduced Time to Capability, TCO

• Virtualized, Resilient, Green Data Center

• Accelerated, Video Enabled Global Network

• Cisco on Cisco Global Data Center

Current state: “traditional DR” (San Jose + RTP)

Target state: Metro-based Virtual DC (Texas/Richadson Metro + RTP/DR)

DC1

(RCDN9) DC2

SJC RTP

(DR)

DC1

(RCDN9) DC2

• RPO = Recovery Point Objective:

Maximum amount of data loss the

business can tolerate (time)

• RTO = Recovery Time Objective:

Maximum downtime the business can

tolerate (time)

RTP

(DR)

RTP

(DR)

• ADL = Acceptable Data Loss (like RPO,

but only for incidents limited to single

facility)

• ART = Acceptable Recovery Time (like

RTO, but only for incidents limited to

single facility)

• RPO = (same as traditional)

• RTO = (same as traditional)

NEW

Operational

continuity

Disaster

Recovery

Richardson Data Center

Richardson Campus Site Map

• GDCP Strategy

Transformation

Growth

Resiliency

• Enablers

Intelligent Network Infrastructure

New Technology

New Services

Footprint for Future

7

Electrical control room-

Station#2

Generator & fuel tank yard-

Station#3

Cooling

towers

TST-

Station#4

Chiller

room

Build room-

Station#5

Data Hall 1-

Station#6

Data Hall 2

Data Hall 4 Data Hall 3

RCDN9 lobby-

Station#1

Gallery-

Station#7

CD1

EF1, EF2

located 3rd fl.

CD2 UPS-A

Distribution

UPS-B

Distribution

9

Electrical control room

Generator & fuel tank yard

Cooling

towers

TST

Chiller

room

Build room Data Hall 1

Data Hall 2

Data Hall 4 Data Hall 3

Main Switchboard B

Main Switchboard A

Cooling Towers

1000 Ton Capacity (ea)

Fuel Tanks

48,000 gals

Chillers

1000 Ton Capacity (ea)

Meter/Switchgear

Generators

Generators

Paralleling Switchgear A

Paralleling Switchgear B

Cooling Tower

Pumps

11

Switchgear A1 Switchgear A2

Switchgear B1 Switchgear B2 UPS B2

UPS B1

UPS A1 UPS A2

Batteries

Batteries

IT storage room

IT/Facility control

rm & office space

2nd floor CUP

Generators

Critical Load

(Servers)

MSB-A

UPS-A

RPP-A

MSB-B

UPS-B

RPP-B

A B

Tie

Breaker

Utility - A Utility - B

40%

max

load

40%

max

load

19

“TD New”

Nexus 7010

UCS - MVDC ONS - DWDM Nexus 7018

• ONS Gear

• Supports DWDM for MVDC

• Enables Resiliency for Data Replication

• Located in DH1 and DH2 for NPRD MVDC

• Nexus 7018

• Provide Scalability

• Supports new UCS Server Bays

• Supports new Teradata

• Supports MVDC

• UCS clusters

• NPROD

• MVDC NPRD

• UCS PROD

(1) Server Bay = 2 clusters or 80 blades

• TD New Node

• Support Cisco Data Warehouse

• Installed in DH 1

• Physical to Virtual

• Increased Density

• Improved Power Utilization

• Lower unit cost

• Transformation to Cloud (Unified Management)

© 2010 Cisco and/or its affiliates. All rights reserved. 31

A tremendous step forward in technology

New processor architecture

New platform architecture

New memory subsystem

New I/O subsystem

New options with SSDs

An even bigger step forward for IT capability

Performance Energy Efficiency

Virtualization


Typical System

Either • 12 DIMMs @ 1066MHz

• Max 96GB (192 GB @800MHz QR)

Or • 18 DIMMs @ 800MHz

• Max 144GB

at lower performance

Intel Xeon 5500 Series with UCS • 48 DIMMs @ 1066MHz

• Max 384GB per Blade

at full performance

Benefit •2x capacity

•Lower costs

•Standards DIMMs, CPUs, OS

Typical Memory Cisco UCS Memory

Xeon 5500 Fixed

number of

DIMMs

can be

addressed

by the CPU

Each DIMM

the CPU

looks for

is made of 4

standard

DIMMs

Xeon 5500

UCS uses significantly fewer cables

Traditional blade server,

similarly configured to… Cisco Unified Computing

March 2009

40% Savings on Cable Infrastructure

Notes: Assumes pre-UCS average V2P ratio of 10 to 1 and post UCS average ratio of 20 to 1

due to the memory expansion technology. Unified Fabric efficiency gains result from power

optimization. UCS efficiency gains result from additional power benefits of UCS.

Cisco IT’s Case Study– Summary

Traditional Unified Fabric

UCS

DC efficiency 100% 130-150% 130% 170-200%

10,000 sq ft, 1 MW

Cabling

$2.7 million $1.6 million $1.6 million

Physical Server Count per Footprint

720 930 -1080 1200-1400

VM Count 7200 9300-10800 12000-28000

Density

efficiencies

12,000 to

28,000

VMs -- In

the same

size DC!

~40%

Savings

from

cabling

March 2009


Component

Management

Interfaces

Cisco (320 servers in 40 chassis

across 6 racks)

HP (320 servers in 20 chassis

across 6 racks)

No. of Chassis

Manager Interfaces 0 20

(OA)

No. of Chassis

Interconnect Interfaces 0 5

(VC Manager)

No. of External Switch

Interfaces 1

(UCS Manager)

4 (3rd Party ToR Switch)

Total Management

Interfaces 1 29

*External switches are required for both Cisco and HP in order for any blade server to communicate with any other blade server.


Embedded Management

B1

Disjointed Management

VM

Manager

iLO

Licenses

Rapid

Deployment Virtual

Connect

Enterprise

Insight

Dynamics

Onboard

Admin –

CMM

Insight

Recovery Insight

Control

California Fabric

Manager Architecture Existing HP Mgmt Architectures

Servers Network

Storage

Access

Embedded

S/W

Embedded Device Management

Multiple Management Solutions

Uncoordinated / Complex Control Plane Integration

Management abstracted to Fabric Interconnects

Chassis have no state – just hold blades

All blades share same management domain


Server

Blades

Adapters

Chassis

Modules

Multi

Chassis

Access

Layer

FC

Enet

FC

Enet

Unified Fabric

Unified Fabric

Unified Fabric

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

• VIRTUALIZE, VIRTUALIZE, VIRTUALIZE

• ESX Farms built using UCS clusters (2 Viffs per cluster)

• Farms built using 192GB as a standard

• Provisioning virtual hosts via SRT/CITEIS

• Migrating SAN to NAS for ESX farms (evaluation)

• Bare Metal automation - next

• Storage Provisioning/Virtualization

UCS 3 Rack Design Initial design built out

with UCS 6120’s.

Place for easy upgrade

to UCS 6140’s.

Cisco 3845 for OOB management of

fabric interconnects

Initial design

supports 5

chassis per

cluster,

increasing to 10

chassis with

6140’s

Each chassis provides

80GB/s throughput

configured with 3m and

5m twin-ax.

Fiber available for

longer runs (legacy

DC)

Anticipated

power

consumption

per chassis:

2.5kw


Cisco 3845 for OOB management

of fabric interconnects

UCS Single

Rack Design Initial design built out

with UCS 6120’s.

Place for easy upgrade

to UCS 6140’s.

Initial design

supports 5 chassis

per cluster,

increasing to 10

chassis with 6140’s

Anticipated power

consumption per

UCS: 6kw to 10kw

Anticipated power

consumption per

chassis: 2.5kw

Nexus 70xx

• Provide Scalability

Nexus 7010 = 256 10G ports compared to Catalyst 6509 = 112 10G ports

Nexus 7018 = 512 10G ports compared to Catalyst 6513 = 112 10G ports

• GREEN

Catalyst 6500 16-port 10G line card: 28.6W per port

Nexus 7000 32-port 10G line card: 23.4W per port

• Supports UCS Server Bays (PROD/NPRD/DMZ)

• Supports Data Warehouse

• Supports Application Migrations

Nexus


Four

Nine

Thirteen


We migrated 462 apps to UCS platform to date.

By the end of Q2 FY11, the number will be around 487.

The Key applications :

Cisco Channels Release Management

Cisco Network Academy

Cisco Tele-presence

Cisco Internal Directory

Cisco Internal Learning (e-Learning)


Key Advantages

Simplified management. UCS & Linux

OS management is simpler & more

automated

~75% Reduction in overall power

consumed

>$600K reduction in hardware/OS cost

Customer X Pre UCS:

HP Superdome 32 shared for

prod & test

• Prod nPar: 20 CPU cores, 48

GB RAM

• Test nPar: 2 CPU cores, 8 GB

RAM

Customer X on UCS:

• Prod: 1 B440M1 server

• 32 CPU cores, 256 GB RAM

• Test: GB B200-M1 blade

• 8 CPU cores, 96 GB RAM Increased CPU cores & Memory

using less power

Performance increase as we adopt

latest Intel technology

Cisco on Cisco. UCS servers capable

of hosting previously Unix-only high-

performance applications

Cisco Texas DC2

James Cribari


July 2009

July 2010

TXDC2 was handed over by

WPR on Dec. 1, 2010


• Cisco multi-tenant, high density data center

• Total area – 162,000 Sq Ft

• Data floor – 27,000 Sq Ft (spread over two halls). No raised floor

• LED lighting in Data Hall; 100kW photovoltaic system

• 2N Systems: Primary Utilities, Generators, UPS; N+2: Chillers

• Building hardened to wind speed requirement of 175 mph

• Goal: Gold LEED Certified Data Center

• Geothermal wells

• Active-Active DC (Two 10 Gb dark fiber connections to RCDN-9 DC)


Category Legacy Design Allen Design

Inlet Temperature 68-72F (20-22C) 78F (25.5C)

Air separation None Hot/cold air separation

based on chimneys

Cooling CHW System with

CRAH units

Airside economizer

(free cooling)

Density 1.5-4.5 kW/rack Average 10.5 kW/rack

Up to 18 kW/rack

Power distribution 480V Back of House

208/120V to Rack

4160V Back of House

415/240V to Rack

UPS Systems Static Dynamic (flywheel)

Certification n/a LEED

PUE Target 2.0 1.34 annual (1.28 – 1.41)


• Stand-alone DC facility

• Separated from RCDN9 DC by around 20 miles

• 5 MW IT load (with additional 5MW possible in future)

• 200 watt / sf density yielding approximately 25K sf of data center floor

• Tier 3 per the Uptime Institute


• 5.5 MW of critical load capacity (potential expansion to 11 MW in future – requiring additional buildout)

• 2N electrical service

• 2N UPS – dynamic/flywheel system

• 2N Generators (approx 96 hr run time)

• 4160V Distribution System

• 230V cabinet power distribution

• All IT equipment to be dual corded

• Power delivered via overhead busway system with in-rack smart CDUs (power strips)


• Hot air discharge containment via chimneys for all cabinets

• IT equipment operational target of 78 F inlet supply air temperature

• AHU capacity is based on an average 25 degree delta T across the IT equipment

• Overhead air delivery

• Airside economizer system

• N+2 chillers and cooling towers, N+1 pumps

• 15 minute thermal storage tank

• 48 hrs Domestic water storage


• 2 diverse data ductbanks

• No raised access floor

• No suspended lay-in ceiling – exposed structure and equipment

• Overhead delivery of power and network

• Building hardened to 175 mph

• Vestibule Entry, no security turnstiles

• Layout will be designed to include a Customer Tour route and viewing panels into the Data Center & Building System rooms

• Target PUE of 1.3 0r lower

• Office space with approximately 20 workstations and areas to support customer visits


• LEED NC v2.2 – target is Gold certification and current indications are that we will achieve this

• Utilize airside economizer systems

• Utilize elevated IT equipment inlet temperatures (78F)

• Implement Cisco Energy Wise to provide efficient facility energy management

• Utilize higher voltage power distribution

• Utilize dynamic UPS systems

• Solar cells on roof space to reduce utilities consumption

• Ground source heat pump systems for office area HVAC


Overhead Delivery / Partial Containment


First Floor Scheme

Global Networks and Data Centers Briefing...Catalyst 6509 = 112 10G ports Nexus 7018 = 512 10G ports...

Documents

Transcript of Global Networks and Data Centers Briefing...Catalyst 6509 = 112 10G ports Nexus 7018 = 512 10G ports...