Title: GPN Disaster Recovery Seminar September 2526, 2007
1GPN Disaster Recovery Seminar September 25-26,
2007
- Scott Silverstein
- Director, Channel and Enterprise sales
- ssilvers_at_ciena.com
- 303-302-3450
2Agenda Day 1
- 100 200 BC/DR / COOP Making sure you are
prepared Listen as a outside consultant
share the common Mythconceptions of BC/DR - 200 300 SAN 101 An overview of storage
technologies and protocols, differences - 300 315 Break
- 315 415 GPN Panel Discussion Campus Issues,
Capital Issues, Data Center Constraints - 530 730 Share refreshments with other GPN
members and our sponsoring vendors - Sullivan's Steakhouse Wine Cellar
3Agenda Day 2
- 900 - 1000 A wide selection of products and
services How do you Choose? IBM will
detail their product offerings and BC/DR
solutions - 1000 1100 Consolidating multiple protocols
and applications - Intelligence on the edge
of your network - 1100 1200 Business Protection Services
Qwest will detail how to protect your business
in the six most critical areas - 1200 100 Break for lunch
- 100 230 Panel w/ round table discussion
- 245 Conference Adjourns
4Who is Ciena?The network specialist
Serving the worlds largest and most advanced
service provider, government, and enterprise
networks
- Specialty Practical innovation for creating
software-defined, service-selectable networks for
highly flexible, adaptable, manageable and
assured networking - Key offerings
- Optical, data and access networking platforms
- Network and service management systems
- Global network services and professional
services - Foundation for critical networks worldwide
- Two-thirds of the worlds 25 largest service
providers - Global 2000 enterprises in finance, retail and
healthcare - Federal, state and local government agencies
- Founded 1992
- HQ in Maryland USA
- Cumulative global market leader in long haul DWDM
and Intelligent Optical Switching - Over 250 Patents held with 100 pending.
- Numerous RD centers worldwide.
- Service and support across the Americas, EMEA and
Asia
5Differentiating BC DR
- Business Continuity
- A Concept or plan that proactively seeks to put
in place process, policy, systems and evaluation
methods to avoid a down condition. - Disaster Recovery
- A subset of BC that deals with the actual
reactive process in the event of a down
condition.
6Business Continuity Requires Geographic Diversity
Power grid diversity
7Can it happen to you?
- During the WTC attack, Pace University was hit
hard. According to Frank J. Monaco, CIO of the
university Pace lost dozens of people during the
attack. Its World Trade Institute, which occupied
the entire 55 floor of the WTC Tower One, was
completely destroyed. The main New York City
campus, which is less than three blocks from the
WTC site, was totally paralyzed with no public
phones, dorm phones, or Internet connectivity.
The lives of over 14,000 students and 2,500
staff, faculty were seriously interrupted with no
academic activities or school services for about
7 days. - The LSU Hurricane Public Health Center (HPHC),
also known as the Center for the Study of Public
Health Impacts of Hurricanes, began
multidisciplinary research on public health
aspects of hurricanes and major floods in 2002.
Comprised of a growing list of over forty
researchers and advisory board members, in fields
ranging from engineering to computer modeling to
medicine, the initial pilot study of the Center
was to assess and mitigate the public health
impacts of a major hurricane strike in New
Orleans. This catastrophic event was in many
ways realized in the fourth year of the five year
project - by Hurricane Katrina in August
2005.Prior to hurricane season 2005 however,
many research initiatives from the Center were
able to be applied to pre-Katrina hurricane
planning and mitigation, as well as later
response and recovery efforts
8A New Budget Market
Emergency Management is a New Budget Market.
Homeland Security funding is now being allocated
to all-hazards not exclusively to terrorist
threats.
New in FY2006 1.8 Billion in funding to 35
identified high-threat regions.
- Hurricanes
- Pandemics
- WMDs
- Earthquakes
- Firestorms
- Landslides
- Tornados
- Flooding
Source Center for Digital Government DHS FY 06
9Current Threats
- Coastal storms account for 71 of recent US
disaster losses annually.
- Annual damage from tornadoes, hurricanes, and
floods is 11.4 billion a year.
- Drought results in average annual losses between
6-8 billion in all sectors.
- Over 10,000 dams in the US classified as
high- hazard 10 of all dams in the country.
- Source FEMA, NOAA Magazine, National Center for
Atmospheric Research (NCAR), H. John Heinz III
Center for Science Economics and the Environment
10Emergency Preparedness Response Have you
thought of everything?
Natural Disasters Severe Weather Earthquakes
Extreme Heat Floods Hurricanes Landslides
Mudslides Tornadoes Tsunamis Volcanoes
Wildfires Winter Weather
Bioterrorism Anthrax Plague QFever Tularemia
Recent Outbreaks and Incidents Bridge
Collapse Anthrax Salmonella Formaldehyde Asbestos
Botulism Blast Injuries Salmonella Wandsworth
Outbreak XDR Tuberculosis Acanthamoeba
Infection Salmonella Outbreak Avian Flu
http//emergency.cdc.gov/
11Im 99 Sure Weve Thought of Everything!
- How good is 99
- 120 Newborns would be given to the wrong parents
each day - 30,560 copies of tomorrows Wall Street Journal
would be missing one of the three sections - 200,000 incorrect drug prescriptions would be
written each year - 8,800,000 credit cards would have incorrect
cardholder information on there magnetic strips
12BC/DRCommon MythConceptions
- Bill Bedsole
- President
- William Travis Group
- bbedsole_at_williamtravisgroup.com
- 847-303-0055
13Storage 101Bringing Up SAN
- Garry Moreau Senior Staff Alliance Consultant
- Ciena Communications
- gmoreau_at_ciena.com
- (763) 421-1449
14DAS Direct Attached Storage
- DAS
- Storage is captive behind server CPU
- Data Access is file system and platform dependant
- Server CPU must handle user I/O requests
- Costly to scale, complex to manage
Intranet
15NAS Network Attached Storage
- NAS
- File Oriented Access
- Multiple Clients
- Shared Access to Data
16SAN Storage Area Network
- SAN
- Block Level I/O Access
- Interconnects different types of storage devices
- SANs can incorporate subnetworks with
network-attached storage (NAS) systems
17Determining BC/DR Methods
- Recovery Time Objective (RTO)
- Time needed to recover from a disaster
- How long can you afford to be without your
systems - Recovery Point Objective (RPO)
- Age of the data you want the ability to restore
in the event of a disaster - Any data created or modified inside your recovery
point objective will be either lost or must be
recreated during a recovery - Network Recovery Objective (NRO)
- Time required to recover or fail over network
operations
18Disaster Recovery Concepts
Recovery Point
Recovery Time
- Recovery Point Objective (RPO)
- Point in time to which applications data must be
recovered to resume business transactions - Recovery Time Objective (RTO)
- Maximum elapsed time required to complete
recovery of application data
19Recovery Point Objective (RPO)
- Business needs drive the technology choice
Recovery Point
Recovery Time
20Recovery Time Objective (RTO)
- Business needs drive the technology choice
-
Recovery Point
Recovery Time
- Recovery Time includes
- Fault detection
- Recovering data
- Bringing apps back online
-
21Synchronous Mirroring
DS3, OC-3/12/48 STM-1/4/16
MAN WAN
OR
OC-3/12/48 STM-1/4/16
Source
Target
22Asynchronous Mirroring
DS3, OC-3/12/48 STM-1/4/16
MAN WAN
OR
CN 2000
OC-3/12/48 STM-1/4/16
Target
Source
23GPN Panel DiscussionWhat are the issues?
Delmar R Johnson South Dakota State University
- Greg Monaco, Ph.D., Executive Director
- Great Plains Network Consortium
24Business Continuity Recovery ServicesKeep your
business up and running with affordable solutions
- Karen Watzel
- Sales Program ManagerBusiness Continuity and
Resiliency Services - IBM
- Klwatzel_at_us.ibm.com
- 719 325 7664
25Consolidating Multiple Applications
ProtocolsIntelligence on the edge of your
networkMAN/WAN Infrastructure for BC/DR
- Garry Moreau Senior Staff Alliance Consultant
- Ciena Communications
- gmoreau_at_ciena.com
- (763) 421-1449
26AGENDA
- Ciena Deployment Overview
- Networking Options
- Performance Considerations
- Deployment Examples
- Summary
27Application Solution Qualifications
Qualifications
Interoperability
28Ciena Data Center and Optical WAN
SolutionsEnterprise RE Government
Applications
Regional and National Backbones Optical
switching transport for government agencies and
enterprises
29Ciena Solutions for IBM Mainframe Environments
Remote Data Center
- Lowest networking costs for GDPS / STP
- Only IBM-certified system capable of multiplexing
STP channels onto a single wavelength - This in turn drives highest scalability
- Unprecedented flexibility, with programmable
approach to enable evolution to advanced protocol
support (IB)
Long Distance Storage and LAN Extension For Metro
Mirror, Global Mirror, z/OS Global Mirror,
Metro/Global Mirror, SVC, PtP-VTS
Up to 1,000s of km
WDM For GDPS/STP Metro Mirror, Global Mirror,
z/OS Global Mirror,
WDM over dark fiber
Primary Data Center
Remote Data Center
Up to 100s of km
30Cienas Enterprise Product Portfolio
ON-Center Manager
CN 2000
CN 4200 MC
CN 4200
31AGENDA
- Overview
- Networking Options
- Performance Considerations
- Deployment Examples
- Summary
32Choices Connecting Data Centers
- Deploy Fiber between buildings
- Build your own network (likely using WDM
equipment) - Lease a connection from a carrier or service
provider - Likely a SONET service, wavelength service, or
an IP service - Manage your own network, and the gateways
- Lease a Managed Storage Service
- Combination of the gateway solution the
connectivity - Customized SLAs and service features focused on
extension
33Wide Area Network Considerations
- Cost
- Must meet Budget Constraints
- Security
- Guaranteed Isolation of Sensitive Data
- Guaranteed Data Delivery
- Performance
- Minimal Impact on the Application with a High
Throughput, Low Latency, and Rapid Restore Times
- Capacity
- Intelligent utilization of network resources
- Manageability
- Ability to Monitor/Report/Protect to Maximize
Performance and Perform Rapid Fault Isolation. - Flexibility
- Support for all Data Types (Storage, Voice, Data,
Video) and Applications
34MAN / WAN Options
Available Transports
- WDM / Dark Fiber Multiple services on a fiber,
high capacity, campus or metro - SONET/SDH Secure bandwidth, ubiquitous service
access - GbEQoS point-to-point GbE private line services
- IP/PoSStorage mapped into IP packets, usually
through TCP/IP - ATM/PoSUsed by legacy channel extension
technology, costly
35MAN/WAN Options
36IP Routed Network
GbE
GbE
FC
FC
GbE
GbE
FC
FC
FC over IP
FC Switch
FC over IP
Switch
FC Switch
Switch
Router
Router
- Storage extension over IP networks (e.g. a leased
GbE service) is only feasible if very high QoS
SLAs can be guaranteed - Latency less than 5 ms
- Packet Delivery Ratio (PDR) of 99.99 or higher
- TCP can drive a maximum of 100 Mbps over a
network with a latency of 5 ms and PDR of 99.99 - IP networks must be dedicated to SAN extension in
order to maximize PDR and minimize latency and
even then, they can not support synchronous
applications due to their high latency - Bandwidth must be over-provisioned by up to 50
to account for dropped packets and retransmissions
37ATM Dedicated Network
ATM Network
OC-3c
OC-12c
FC
FC
OC-3c
OC-12c
FC
FC
Channel Extender
FC Switch
Channel Extender
ATM Switch
FC Switch
ATM Switch
- Channel extenders represent a legacy approach to
storage extension - Support ESCON/FC/FICON neglecting overall data
center extension requirement likely including GbE - ATM mapping adds significant overhead resulting
in bandwidth inefficiency - Extremely high pricing
- Adds significant latency to applications not
requiring host emulation resulting in decreased
application performance
38Gigabit Ethernet Private Line Services
- GbE over wavelengths
- Mapping directly into wavelength
- Dark fiber distances
Fiber
GbE
GbE
DWDM
DWDM
- GbE over SONET/SDH
- Mapping directly into TDM
- 43 of Ethernet Services in NA are based on SONET
GbE
GbE
SONET SDH
SONET SDH
SONET SDH
- GbE over switched service
- Over shared service
- Similar performance to IP VPN service
GbE or 100bT
L2 Switched Ethernet Network Or L2 Ethernet VLAN
over L3 Network
Carriers GbE services can have high performance,
BUT still need to consider the switching/routing
equipment within an enterprise
39Storage over WDM
Protocol Rate
Service
Fiber or Wavelength
Characteristics for Storage
Todays Applications
- Fiber relief
- Native protocol carriage (GbE, Fibre
Channel, ESCON) - Virtual Private Network
- Bandwidth Leasing
- Large capacity (80G in Metro)
- Reach limited to metro distances
- DWDM is cost effective for large of services
and volumes of data (gt 10 Gbps) - CWDM provides a lower cost solution
40Storage over SONET / SDH
- What is SONET / SDH?
- Self-monitored high performance networking
technology - Ubiquitous network with over 150,000 installed
carrier rings - Uses Time Division Multiplexing (TDM) to
aggregate multiple signals together - Standardized rates from 50Mbps 40 Gbps
- Why SONET / SDH?
- Guaranteed, high bandwidth
- Low latency
- Deterministic
- Secure, 99.999 availability network
- Metro and Long Haul networks
- National and International
SONET/SDH Based Services Perfectly Match the
Requirements of Business Continuance Applications
41SONET as the Backbone
Native Services are Transported Worldwide over
SONET
- 99 of all data traffic goes across the SONET
network, including IP - Native protocols are mapped directly into SONET
as soon as they leave the building or campus
where they originated
42AGENDA
- Overview
- Networking Options
- Performance Considerations
- Deployment Examples
- Summary
43Performance Considerations
- Latency
- Delay introduced by the intermediate equipment
(i.e. switches, transport, speed of light in
fiber, packet loss, jitter) which slows down the
response time of applications - Bandwidth
- Dedicate enough bandwidth to ensure optimum
application performance, but only allocate the
required to bandwidth to minimize MAN/WAN costs - Choose a storage networking technology that
efficiently makes use of expensive MAN / WAN
resources - Protocol Flow Control
- Not have sufficient protocol extension
capabilities will leave the application waiting
Latency, Bandwidth and Protocol Flow Controlcan
significantly impact application performance
44Performance Considerations Network Latency
Element Switching
- Time of Flight
- Time it takes light to traverse the network
- 5 ?sec per km
- 1 ms per 125 miles
Data Center
Data Center
45TCP/IP Flow Control
A packet is missing, please resend the last 20
packets
Lost packet
Enterprise Location 1
Enterprise Location 2
- TCP enables guaranteed services across lossy IP
networks - TCP has a mechanism to request retransmission of
lost packets - Bandwidth wasted increases with high data rates
and latency - When extending high-end applications over a lossy
metropolitan or wide area network, severe
performance limitations emerge - Need to consider
- Packet Delivery Ratio
- Packet Error Ratio
- of TCP Retransmits
Sources of Inefficiencies
- Time to recognize that a packet is lost
- Time for lost packet message to be sent
- Amount of data resent
Storage over IP devices typically offload this
retransmission requirement from the storage, but
this does not eliminate the end to end latency
46Packet / Cell Loss Contributions
IP Network
Typical operating region
Required operating region
Provisioned Bandwidth
1200
1000
800
Throughput
Average Data Throughput
Mbps
600
400
Lost Packet
200
Timeout
0
99
99.99
99.9
99.999
- Lost packet results in cutting throughput in half
- Throughput slowly recovers over time
- Similar effects with other IP and ATM Storage
Solutions
Packet / Cell Delivery Ratio
Due to lost packets and retransmissions, goodput
is artificially limited
Performance of Storage Across IP and ATM
Infrastructures is Significantly Affected by Loss
47Performance Considerations BandwidthTradeoff
Between Cost and Application Performance
- Cost
- MAN/WAN networking can be greater than 60 of the
BC/DR application cost - Minimize the bandwidth used through
- Data Compression
- Efficient Mapping
- Application Consolidation over a single transport
- Performance
- Need to consider the actual throughput of data
- Need to guarantee application isolation and
efficient delivery of data - If not enough bandwidth is allocated, or too much
contention in the network, a bottleneck to the
application will occur
Optimum SolutionBandwidth Allocated
Application Requirement
48Typical Deployment Comparison
- Option 1 Storage over IP Routed Network
May be consolidated into a single device
DWDM/OC-3 or GbE
FCIP
FC/IP
FC
FCIP
FC/IP
FC
FC
FC
MAN/WAN
Optional Switch
FCIP Gateway
LAN Switch
FCIP Gateway
Router
FC Switch
Router
FC Switch
Disk
Disk
LAN
LAN
Multiple G/Ws, switches and routers add
latency, dropped packets, lower throughput and
complexity (Both Sides)
Option 2 Storage over DWDM/SONET/GbE Network
FC Pt-to-Pt mode
FC Pt-to-Pt mode
DWDM/OC-n or GbE
or
or
FC
FC
FC Fabric mode
FC Fabric mode
MAN/WAN
GbE
Disk
GbE
CN2000 Storage over SONET/DS3
Disk
CN2000 Storage over SONET/DS3
FC Switch
FC Switch
LAN
LAN
Router
Guarantees layer 1 isolation, yet aggregation and
data compression of all traffic (FCLAN) before
connecting to the WAN
49Data Compression
Protocol Line Rate
Application Rate (varies)
Compressed Data Rate (dependent upon application
content)
1 Gbps
300 Mbps
lt150 Mbps
Compressor
Active Multiplexing
Data Frame
Data Compression
IDLE Frame
Content Compression IDLE removal
- Each port has dedicated compression chips,
ensuring wirespeed performance and maximum
compression - Uses industry standard LZS (Lempel Ziv Stac)
algorithm - Automatically disabled on a per-packet basis when
data not compressible - Available on all protocols and all interfaces
- FC/FICON/ESCON/GbE
- DS3, OC-3/12/48, STM-1/4/16, DWDM, Dark Fiber
Per Port Hardware Compression Dramatically
Reduces WAN Bandwidth Costs
50Reducing BandwidthDynamic Bandwidth Sharing
Remote Data Center
Disk Mirroring
Fibre Channel
65
Aggregated Traffic
Tape Backup
GbE
20
15
GbE
LAN
- Each port is assigned a guaranteed minimum
percentage of the bandwidth - The percentage can be assigned to any value from
0 to 100 - When only one port has traffic to send, it has
access to the entire link
51Performance Considerations Standards Protocol
Efficiency
52AGENDA
- Overview
- Networking Options
- Performance Considerations
- Deployment Examples
- Summary
53Certification Test ResultsIBM Global Mirroring
in IBM Tucson
Global Mirror Establish Bandwidth
Ciena
240
200
Ciena
160
Competitor
Bandwidth
(MB/sec)
120
Competitor
80
40
0
0
2000
4000
6000
8000
10000
12000
PPRC Distance
PPRC Distance
(miles)
(miles)
- Data was extracted from an IBM white paper
titled - IBM TotalStorage ESS Global Mirror
- Asynchronous Peer-to-Peer Remote Copy Performance
Perspectives - Aug 2004
54Deployment ExampleStorage over SONET
Why Customer Choose Storage over SONET
- gt 50 capital cost savings
- 2X improvement in WAN utilization
- Detailed performance monitoring of services and
WAN
55Deployment ExampleFortune 1000 Financial Services
Atlanta, GA
Sterling Forest, NJ
Optional LAN traffic GbE
Carrier SONET Service
56Deployment ExampleDWDM and Storage over SONET
Why the customer choose DWDM Storage over SONET
- DWDM provided extremely low latency and
significant capacity - 2.5 Gbps today scaling to 80 Gbps
- Leased a SONET OC-3 service as a backup
57AGENDA
- Overview
- Networking Options
- Performance Considerations
- Deployment Examples
- Summary
58Summary
- Costs due to WAN extension dominate the business
case for BC/DR solutions - Solutions vary greatly based on cost,
performance, security and data availability
targets there is no single answer - Carriers and Service providers must be leveraged
for solutions beyond fiber only - New products are available that both simplify the
deployment of, and help drive down the costs
associated with distance solutions for storage
extension
59Planning for the WorstNetwork Design
Considerations
- Rick Spencer
- Staff of the CTO
- Qwest Communication
- rick.spencer_at_qwest.com
- 720-578-9198