Transcript How to sell Networking

IBM System Networking
Anthony Angell – Solution Sales BUE, North America
1
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Agenda
 Basics
 SDN
 Cloud Competition
 Portfolio – building a configuration
 Selling Converged Systems today
2
This educational material is intended for your use in selling.
© 2014 IBM Corporation
What: Illudium Q-36 Explosive Space Modulator
Why: Blow up Earth
3
This educational material is intended for your use in selling.
© 2014 IBM Corporation
“The Line”
Core Networking
Row/Rack Networking












Monolithic
Entrenched
Proprietary
Expensive
Irreducible
Hardware dependencies
Single Vendor Market Domination
Distributed
Rapid transition to Merchant Silicon
Pooled
Rapid software incursion
No significant hardware
dependencies
 Multi vendor, particularly converged
Other ‘Core Networking like” products include Other “Row like” products include x86
IBM Mainframe and EMC VNX/Symmetrix
servers and local storage
4
This educational material is intended for your use in selling.
© 2014 IBM Corporation
5
Mark Andreessen: Founder
– Netscape and Andreessen Horowitz© 2014 IBM Corporation
This educational material is intended for your use in selling.
SDN – Software Eats Networking
What:
 Incorporated into client’s existing Hypervisor install base
• VMWare’s NSX
• Microsoft’s Windows Server with Hyper V
• OpenStack Neutron (formerly Quantum)*
 Takes functionality from the network hardware and moves it to the Hypervisor
Why:
 Easier to administer
 Feature addition at the speed of Software vs Hardware
 Centralized Network control vs distributed
 Logical vs physical. Resource Pool vs Discrete Allocation
 Cap X reduction
6
This educational material is intended for your use in selling.
* Not hypervisor based
© 2014 IBM Corporation
We also update our bare metal impact analysis in
EPS and find that lower OS pricing has materially
lowered our central case "SDN translated" EPS to
$1.67. A detailed price comparison of bare metal
solutions with Cisco shows a large pricing gap that
we believe will have to close for Cisco to maintain
share,“
Excerpt from JP Morgan Cisco downgrade, 2Q 2014.
Converged Systems
What we’ve been selling
8
© 2014
IBM Corporation
Why clients are buying differently
now
Converged Systems – what makes them different?
 Three Sales in One
• Storage, Networking and Server components all have a sales cycle
• Every vendor has an advantage and a disadvantage in one of the three
• Sales tend to be lost in the blind spots
 Not understanding the “disadvantage” space leads to incorrect configurations
• More than half the designs we see quoted have incorrect Network configs
• Most designs we see are sub-optimal vs. competition
• Workload differentiates Converged Systems – get specific
 If you sell BAU clients will buy BAU – bad news unless you are the incumbent
9
This educational material is intended for your use in selling.
© 2014 IBM Corporation
The Cloud Conundrum
The success of cloud within existing IT organizations could be measured by
the reduction of both Employees and Cap X.
 “New IT or No IT”
 The truth about “non transferable” workloads.
 The rise and rise of Shadow IT.
 Talking points with your clients
• Premium priced hardware vs cloud
• Vendor lock in vs Open/Open Source. “The Wedge”
• “How long…”
10
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Putting it All Together
11
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Data Center Network Evolution
(“No network in the POD”)
Push the network boundary upstream
Ancient History
10+ years
Data Center Network (LAN|SAN)
Began at the Server Edge
Interconnect
Network
Admin
Interconnect
Interconnect
Interconnect
…
Interconnect
Interconnect
Interconnect
Interconnect
Interconnect
Past 5 – 10 years
NOW
Data Center Network (LAN|SAN)
Begins at the Chassis Edge
Data Center Network (LAN|SAN)
Begins at the POD Edge
Network
Admin
Aggregated
Virtualized Host Ports
(server admin)
Chassis Level
System
Interconnect
Module
…
Chassis Level
System
Interconnect
Module
Aggregated
Virtualized Host Ports
POD Level System
Interconnect
Fabric
POD Level System
Interconnect
Fabric
…
Interconnect
Network
12
© 2014 IBM Corporation
Chassis Ethernet – Step 1
13
EN 2092
B22 FEX
SI 4093
EN 4093
1G to servers
1G/10G uplinks
10G to servers
10G uplink - limited
10G to servers
10G/40G uplink
10G to servers
10G/40G uplink
Use
Use
Use
Use
Existing 1G network.
Low performance spec–
management/remote.
Existing 10G network
requires port aggregator
vs switch limited uplink
bandwidth.
Existing 10G network
requires port aggregator
functionality.
Existing 10G network
requires full switch
functionality.
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Chassis Ethernet - Considerations
 Bandwidth
• What type of connectivity does the client need? Switch vs Port Aggregator
• How much outbound/intra-chassis bandwidth do they need…and why you should ask?
• What does their network look like today – vendor and models?
 Mode
• Traditional Network mode – Layer 2/3 switching
• End Host Mode – chassis (or POD) appear to the upstream switch as a node
• Why it’s important
 Key Features of Flex Networking
• Flexible Port Mapping – allows client to freely allocate bandwidth amongst ports
• Feature on Demand – allows client to “open” additional ports on their existing switch
• Chassis switches and port aggregators switch local chassis traffic (exception B22)
14
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Flex System moves VMs faster than Cisco UCS
2.5X faster VM migration with Flex System compared to Cisco UCS
due to 4.8X more backplane throughput with Flex System compared to Cisco UCS
Flex System
Cisco UCS
Max. network flexibility
North / South networking only
NEW: White paper
IBM || BP || Client
Higher throughput allows
more transactions for
business critical applications
in cloud and big data
environments.
15
© 2014 IBM Corporation
Landscape Changes
16
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Chassis Fiber Channel (FC) – Step 2
17
FC 3171
Pass-thru
8G to servers
8G uplinks
FC 3171
FC 5022
Special Case:
8Gb to HBA
8 Gb uplink
8/16Gb to HBA
8/16Gb uplink
CN 4093
Use
Use
Use
SANs managed
upstream that don’t
need a FCF.
Existing 8G networks w/
no upgrade plans.
Existing 8G SAN
targeted for upgrade.
16Gb SANs. Brocade.
Qlogic networks.
This educational material is intended for your use in selling.
8Gb FC
10Gb Ethernet
Flex ports
Uses CNA
© 2014 IBM Corporation
Chassis Fiber Channel - Considerations
 8Gb vs 16GB
• Current state and future plans for FC SAN network?
• 16Gb is an advantage, if the client requires 16Gb.
• 16Gb is a disadvantage if priced against competition’s 8Gb solution– particularly UCS.
 When to use converged (CN4093) vs native (FC 3171/5022) and why
 Storage Constituents
• Talk to the storage teams – the requirement for Server HBAs for FC is unlikely to
come from any other group – and it gives us a competitive advantage.
• What vendor are they using? In cases of Qlogic or Brocade we can provide single
vendor technology – the competition can’t.
 Cisco
• UCS does not offer 16Gb FC, native (non FCoE) FC or HBA SAN connectivity.
18
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Multiple Chassis – Step 3
Any deal that involves two or more chassis needs a way to interconnect the
chassis . If we do not architect the interconnect our competition will.
 The System is incomplete without an Interconnect/Fabric.
 If the System is sold correctly, clients will expect a
recommendation on chassis interconnection.
 The configuration used to connect to the client’s existing network
should be your main area of focus – not the hardware.
 Interconnects are just that – not a new network element. The Line
19
This educational material is intended for your use in selling.
© 2014 IBM Corporation
IBM Flex System Interconnect Fabric
Delivering a simplified Fabric Infrastructure to accelerate deployment with optimum efficiency
What is it?…
 Simple Ethernet Fabric cluster
 Loop free design with no spanning tree
 Scalable multi-chassis, multi-rack interconnect fabric infrastructure
Customer Importance…
 Rapidly deploy additional chassis or racks – configure the Fabric once
 Reduce networking management complexity without compromising performance – manage 1 device vs 20
 95%1 reduction in the number of networking devices required to manage a 9 chassis, 3 rack configuration
 ZERO2 down time for service upgrades
 ZERO3 additional Flex System CAPEX to implement – use existing SI4093 and G8264CS Flex System HW
 Support for Ethernet, iSCSI and FCoE protocols
 Proven interoperability with other vendors core network like Cisco and Juniper
1 G8264CS Switch vs 20 Devices (2 G8264CS’s and 18 SI4093’s) 2 master switch provides in service upgrades to all 18 devices
modules enabled for the fabric infrastructure
1
3
G8264CS and SI4093 are the only
© 2014 IBM Corporation
IBM Flex System Interconnect Fabric
21
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Virtualization POD TM (it’s not trademarked)
A clustered group of Flex Chassis using cutting edge transport and software
interconnects designed for HPC-like performance with full SDN software support.
Components:
 G8332 Fabric connectors with the latest Broadcom Trident chipset.
 SI4093 embedded aggregators
Benefits:
 40G fabric within the cluster – no other vendor offers
 75% reduction in cabling, same bandwidth– 160G on 10G -16 cables, 4 on 40G
 8332 will support VxLAN with a software upgrade – critical for SDN
 With Flex Port Mapping the SI4093’s 40G ports are the same price as 4x10G
22
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Selling Converged and Networking Today
24
•
The days of selling discrete Networking, Servers and Storage are over.
•
Client needs have changed
• Open vs Proprietary – why it matters more now
• Software decoupling from Hardware – what’s next?
• Building an aggregated system to run an overlay – changing philosophy
• Building a bridge to the future state now – this is becoming top priority
•
The Line
• Converged Systems are not, and should not, be tied to any higher layer for
Hardware or Software dependencies.
• Sell to what they have deployed above “The Line” and competitively below
•
Clients are looking for direction as the market thrashes – provide it.
•
Why vs What
This educational material is intended for your use in selling.
© 2014 IBM Corporation
25
This educational material is intended for your use in selling.
© 2014 IBM Corporation
Appendix
26
This educational material is intended for your use in selling.
© 2014 IBM Corporation
* Only supports 2-port adapter, can not support all ports on a 4-port adapters
EN6131 – 40Gb Ethernet Mellanox
IBM Flex System EN4023
10Gb Scalable Switch
(Brocade)
Cisco Nexus B22 Fabric
Extender for IBM Flex
System*
Lead Offerings - Differentiation & Value
EN4091 – 10Gb PassThrough*
CN4093 – 10Gb Int. & 10/40G
Ext. – FCoE & low cost FC
connectivity
EN4093R – 10Gb Int. & 10/40G
Ext. (Multi Modes: L2, L3,
OpenFlow, Easy connect)
SI4093 – 10Gb Int. & 10/40G
Ext. Simple Connectivity
EN2092 – 1Gb Int. & 1/10G Ext.
Unmatched Ethernet Portfolio
Specific Needs - Offerings
Flexible & Scalable
© 2014 IBM Corporation
Specific Needs - Ethernet Offerings
EN4091 10Gb Pass-through
Cisco B22 Fabric Extender
• Those requiring only 2-port connectivity on node*
• Those requiring only 2-port connectivity on node*
• Needing dedicated 1-to-1 10Gb pipes upstream
• Those only wanting an Ethernet offering from Cisco
• SI4093 provides more node ports
• Requires upstream Nexus 5548/5596 or 6000 which
provides all the management
• Lowest acquisition cost 10Gb Ethernet module – but
drastically increases external networking costs
• SI4093 can save you $18-52K per chassis on I/O
10Gb
1/10Gb
costs (factor in trans
& upstream
network)**
Total Ports
Downlinks
Uplinks
14
14
Base System
Total Ports
Base System
10Gb
Downlinks
10Gb
Uplinks
14
8
EN6131 40Gb Switch Module
EN4023 10G Scalable Switch
• For clients implementing with external Brocade VCS
•
Provides full 40Gb end-to-end through the chassis
• Ability to scale ports
•
Based on Mellanox technology
•
Competitive advantage against Cisco and Dell
Total Ports
Base System
10Gb
Downlinks
10Gb
Uplinks
40Gb
Uplinks
14
10
0
Total Ports
Upgrade #1
28
10
2
Upgrade #2
42
14
2
Base System
* Not supported with x222
40Gb
Downlinks
40Gb
Uplinks
14
18
IBM offers a unmatched portfolio of offerings!
© 2014 IBM Corporation
Flexibility – Lower Cost of Ownership
Flexibility – Flexible Port Mapping
Scalability
 Most vendors products support a set
number of ports internal & external
 Ability to reallocate ports
•
•
• Example: Cisco UCS, HP and most Dell
 IBM Features on Demand – buy
upgrade license to turn on additional
ports
-
28
10
2
w/ Upgrade #2
42
14
2
10x1Gb for a 10GbE
4x10Gb for a single 40Gb
 Advantage areas – clients that need
• Example: X222 nodes, or nodes requiring
4-6 adapter ports
• HP and Dell users must by 2nd or 3rd pair
SI4093
40Gb
of Ethernet 10Gb
modules => 10Gb
higher acquisition
Total Ports cost, more internal
external
uplinks
External
uplinks
devices to manage and greater
power consumption
Base module
14
10
0
w/ Upgrade #1
Reallocate ports internal/external
Trade off lower for higher bandwidth ports
Supported on EN2092, SI4093,
• A 4-port compute node – only base module
-
HP and Dell must sell 2x the modules
Cisco UCS can but increase oversubscription
• A 6-port compute node – only the base model
-
HP and Dell must sell 3x the modules
Cisco UCS can but increase oversubscription
• Flex System x222 is supported with the base
• 1Gb today but want 10Gb uplinks in future
- IBM – Features on Demand Upgrade1 or 2
- HP and in&most
cases Dell will buy©new
switch
2014 IBM
Corporation
EN4093R
CN4093
Embedded Reference
This educational material is intended for your use in selling. It is NOT a deliverable for your clients
This educational material is intended for your use in selling.
31
© 2014
IBMIBM
Corporation
© 2014
Corporation
1. Understanding compute node requirements & upstream bandwidth
•
What adapters are you going to have installed in the compute nodes in your chassis?
• 1Gb, 10Gb, 40Gb, CNA (FCoE)? (Can answer more than one)
• 2-port, 4-port, or 6-ports (CN4058)? (Can answer more than one)
• What external links do you want coming out of the chassis?
• 1Gb, 10Gb, 40Gb, FCoE or FC?
1GB a partial chassis (having
10Gb Ethernet
Converged
•Compute
Donode
you plan to deploy only
empty node bays)?
Ethernet
Features
Support 2-port adapter1
Support 4-port
adapter1
Support 6-port of 8- port
adapter1
40Gb
Ethernet
EN2092
SI4093
EN4093R
EN4023
EN4091
B22
CN4093
EN6131
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
Yes
No
No
Yes
Yes
Yes
No
No
Yes
No
Support x222 Node
Yes
Yes
Yes
Yes
No
No
Yes
Flexible Port Mapping2
Yes3
Yes3
Yes3
Yes
No
No
Yes3
No
1Gb uplink
Yes
Option
Option
Option
Option
No
Option
No
10Gb uplink
Option
Yes
Yes
Yes
Yes
Yes
Yes
No
40Gb uplink
No
Optional
Optional
Optional
No
No
Optional
Standard
FCoE Transit
No
Yes
Yes
Yes
Yes
Yes
Yes
FC Break Out
No
No
No
No
No
No
Yes
No
1
Assumes two Ethernet modules in switch bay 1 and 2.
2 Ability to sign ports as required internal/external – allows use of x222 / 4-port adapters with out having to purchase upgrades.
3 Available with GA of IBM Networking OS 7.8.
© 2014 IBM Corporation
2. Understand what the client network requirements (chassis & upstream)
•
•
•
•
•
Want clear separation of System Admin and Network Admin = Transparency
Want integrated Layer 2 or Layer 3 networking?
Want investment protection for Software defined networking (OpenFlow)?
What upstream Cisco Nexus might they want to plug into?
Do they want managed specifically by Cisco Nexus or Brocade VDX network?
Compute node
1GB
Ethernet
10Gb Ethernet
Converged
40Gb
Ethernet
Features
EN2092
SI4093
EN4093R
EN4023
EN4091
B22
CN4093
EN6131
Transparent Mode – No Change to
upstream networking
No
Yes
Yes1
No
Yes
Yes-requires
Nexus
Yes1
No
Transparent Mode – No Change to
upstream networking but VLAN aware
mode
No
Yes
Yes1
No
Yes
No
Yes1
No
Layer 2 Networking
Yes
Yes3
Layer 3 Networking
Yes
No
Yes
Yes
No
No
Yes
Yes
Yes
Yes
No
No
Yes
OpenFlow Mode
No
No
Yes
No
No
No
No
Upstream Nexus 2K
Yes
Yes
Yes
Yes
No
Yes
Upstream Nexus 5/6K
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Upstream Nexus 7K
Yes
Yes
Yes
Yes
Yes
No
Yes
Managed by Cisco Nexus
No
No
No
No
No
Yes
No
No
Managed by Brocade VDX
No
No
No
Yes
No
No
No
No
No
The EN4093R/CN4093 can be configured in easy connect mode – requires configuration but once done it is easy to replicate and provides flexibility for future.
The SI4093 can be more cost effective and not require changes in your network.
© 2014 IBM Corporation
3 The SI4093 has limited Layer 2 feature support
1
2
3. Other possible features to consider
•
•
•
Do you want to build POD/cluster with single IP address?
Do you have environment where VM’s or nodes might co exist in same chassis?
Are you looking to implement vNIC’s for your x86 compute nodes?
•
IBM has basic vNIC2 or more advanced UFP.
Are you interested in have a network that is VM aware and can help simplify VM deployment and mobility?
Do you require multi-tenancy? (Switch partitioning)
Clients like Cisco vPC and interested in link aggregation for performance and availability?
•
•
•
Compute node
1GB
Ethernet
10Gb Ethernet
Features
EN2092
SI4093
EN4093R
EN4023
Build POD – single IP
No
Yes4
No
Yes
Yes
No
Traditional Stacking up to 8
No
No
Yes
No
No
No
Yes5
Keep VLAN traffic in the chassis
Yes
Yes
Yes
Yes
No
No
Yes
vNIC2 Support
Yes
Yes
Yes
Yes
Yes
Yes
Yes
UFP Support
No
2Q146
Yes
No
No
No
Yes
No
VMready – VM aware
Yes
2Q146
Yes
No
No
No
Yes
No
Switch Partitioning
No
Yes
Yes
No
No
No
Yes
No
Link Aggregation
No
No
Yes
No
No
Yes
EN4091
B22
Converged
40Gb
Ethernet
CN4093
EN6131
Yes
4
Similar to stacking but requires external Top of Rack switches (SI4093+G8264CS, EN4023+Brocade VDX, B22+ Cisco Nexus)
Support in Hybrid stack 2 x CN4093 + 2 to 6 EN4093R
6 Support with release of GA5 – by end of 2Q 2014
5
© 2014 IBM Corporation
Decision Tree: Understanding clients requirements
Vendor Lock in
(i.e. Cisco, Juniper etc)
Simple
Ethernet
or
Converged
Ethernet only
Cost &
Simplicity
Cost or
Flexibility
Understand
Clients
flexibility
and
environment
FCoE
Converged
Open to alternatives that
provide Value (i.e. All IBM)
Ethernet only
Cost &
Simplicity
Flexibility
modes
Legend
SI4093
EN4093
CN4093
EN6131
TOR
Cost or
Flexibility
FCoE
Converged
Simply
Ethernet
or
Converged
Flexibility Modes
Networking or SDN
EN4093R
SI4093
Full 40GbE
EN4093R
SI4093
Easy connect
L2,L3 or
OpenFlow
Integrated
FC from
chassis Integrated
FC or
FCoE at
TOR
EN6131
Integrated FC
from chassis
Cisco or
nothing
B22
FEX
Brocade EN4023
VCS
Integrated
FC or
FCoE at
TOR
FCoE Transit to
TOR
CN4093
SI4093
Easy Connect
to Brocade
Or Cisco
TOR
G8316 = 40Gb
TOR
G8264 = 10Gb
G8316/G8332 =
40Gb
CN4093
TOR
G8264 = 10Gb
G8316/G8332 = 40Gb
FCoE
Transit to
TOR
SI4093
TOR
G8264CS
© 2014 IBM Corporation
Why Convergence versus Separate LAN & SAN switching?
 Clients can see major acquisition cost savings through convergence
• Some servers now include 10Gb LAN on Motherboard (LOM) which provide significant savings
I/O Comparison
Separate LAN & SAN
Converged Solution
10GbE standard
8Gb FC = $1,849 **
10GbE Standard
CNA functionality $99**
• Other servers with out the LOM can also see savings
I/O Comparison
Separate LAN & SAN
Converged Solution
10GbE $629***
8Gb FC = $1,849 **
10GbE adapter $629***
CNA functionality $829****
 Clients can see savings by reducing the number of FC SAN switch ports required
• The G8264CS can be up to a 25% savings, compared to separate LAN and SAN switching
• Clients can see further savings by leveraging a transit switch, allowing for more servers per port
IBM Web pricing * Emulex 8Gb FC Dual-port HBA for IBM System x (42D0494) ** Emulex Embedded VFA III FCoE/iSCSI License for IBM System x (90Y5178) *** Emulex Dual Port 10GbE SFP+ VFA
III for IBM System x (95Y3762) **** Emulex VFA III FCoE/iSCSI License for IBM System x (95Y3760) ***** G8264 plus IBM® System Storage® SAN48B-5
36
© 2014 IBM Corporation