1. WHITE P APER
System Networks Drive the Next Generation of Automated,
Dynamic Datacenters
Sponsored by: IBM
Lucinda Borovick
December 2011
EXECUTIVE SUMMARY
www.idc.com
Today's competitive business environment is changing the nature of the datacenter,
driving the need for more flexible, scalable, and cost-effective approaches than ever
before. One way datacenter managers are meeting this need is by adopting cloud-
oriented approaches to computing, which in turn is placing greater demand on the
F.508.935.4015
system network in terms of speed, flexibility, scalability, cost-efficiency, and support
for technologies such as virtualization.
The datacenter network is in the midst of an evolution from a fixed, data-centric,
client/server topology to an application-driven, dynamic network better suited to the
P.508.872.8200
needs of cloud computing and Big Data. By incorporating greater flexibility into the
network, datacenters can improve their agility and better respond to business needs
and support changing business requirements. This is driving an explosion in high-end
networking equipment, and IDC expects that by 2015, 10GbE will represent 61% of
datacenter network port shipments.
Global Headquarters: 5 Speen Street Framingham, MA 01701 USA
IDC believes that a converged network infrastructure, combining server and storage
networks onto a unified 10GbE fabric, is the most appropriate technology path to
meet the requirements for today's high-performance datacenter applications. Having
server, storage, and network resources working in concert not only brings benefits in
terms of agility and scalability but also maximizes capital investments by reducing the
need to invest in duplicate hardware.
These networking environments should be standards based and best of breed. This
enables enterprises to pick and choose the best available and most cost-effective
technologies and provides future upgrade paths without locking them into proprietary
offerings.
IBM's System Networking business and portfolio, recently expanded with the
acquisition of BLADE Network Technologies, are designed to address these
datacenter needs in a modular, standards-based, and interoperable approach. By
combining network intelligence with compute and storage systems resources, IBM is
working to deliver a unified fabric networking infrastructure that addresses the needs
of today's virtualized, cloud-based infrastructure in a manner that is scalable, cost-
effective, and easy to manage and that provides the flexibility to allow businesses to
be more innovative and competitive in the market.

2. SITUATION OVERVIEW
One of the most important changes IDC sees today is the trend for datacenters to
migrate to private and public cloud. Cloud computing provides improved flexibility and
scalability as well as operating cost advantages to organizations, but supporting cloud
environments requires a network with the flexibility to support a changing architecture
that is easily managed and open.
Customer Opportunities
Private/Public Cloud
Cloud computing has arisen over the past several years as a viable option for
enterprise IT organizations looking to reduce costs and increase their operating
efficiency. It includes both public cloud, in which services are provided by a third-party
service provider, and private cloud, in which a cloud infrastructure is built by an
enterprise and used to support in-house operations. In 2010, worldwide public cloud
IT services spending exceeded $21.5 billion, and IDC expects this number to reach
$72.9 billion by 2015.
The emergence of cloud computing is being driven by a variety of factors, including
the ability to pay via a usage-based model without having to overprovision and the
improved ability to deploy new applications and services to the end user in response
to new business requests. A recent IDC study of IT managers indicated that more
than half expect private cloud to reduce the cost of their server, storage, and network
hardware, while a significant percentage also expect it to help them better match IT
costs to business needs and to improve application performance (see Figure 1).
FIGURE 1
Expected Private Cloud Benefits
Reduced cost of
server/storage/network hardware
Better match IT costs to business
needs
Improved application perf ormance
Standardization and stability of
inf rastructure
Signif icant improvement of staf f
productivity
0 10 20 30 40 50 60
(% of respondents)
Source: IDC's Private Cloud Management Survey, September 2010
2 #232049 ©2011 IDC

3. Big Data
Another important trend in today's datacenter is what IDC refers to as "Big Data." Big
Data is more than simply a reference to the large and ever-increasing amount of data
being stored in the datacenter — in 2010, over 16,000 petabytes (PB) were shipped
into the market. The real question is what organizations are doing with this capacity.
Under the Big Data paradigm, organizations are constantly updating information
about their customers, prospects, suppliers, and inventory — and acting on it. Data is
no longer simply being stored and forgotten; rather, data of all types is being used to
make proactive business decisions on a daily basis. Supporting this paradigm
requires a fundamental shift in the way data is stored and managed by organizations
and requires powerful real-time data analytics and tools, decision support systems
and dashboards, and automated hooks into the applications that run the business,
such as ERP, CRM, and financial systems.
Supporting Big Data requires fundamental changes in the compute and storage
infrastructure to provide the necessary resources to make these decisions in real
time. It also requires fundamental changes in the network to support the required
levels of throughput, intelligence, and agility.
Maximizing the opportunity for real-time analytics requires an organization to have the
ability to process massive amounts of diverse data and have it move quickly and
efficiently throughout the network. Network performance is imperative with both
consistent ultra-low latency and line rate performance.
Virtualization
Virtualization is one of the key strategies that IT teams are pursuing as part of
corporate efforts to reduce capital costs through aggressive consolidation while
boosting asset utilization, and it is a key building block required for the adoption of
cloud computing. This trend is well under way; in fact, IDC believes that in 2009, for
the first time, more new application instances were deployed on virtual machines
(VMs) than on dedicated physical servers. IDC expects that the virtualization trend
will only accelerate and that the majority of all installed server applications will be
running as virtual machines.
Customer Challenges
Not Optimized for Cloud/Big Data
In today's datacenters, most individual components are not optimized to work
together. In most organizations, storage, server, and network resources are contained
in separate silos, and these architectures are inefficient and not well suited to support
cloud computing, virtualization, or Big Data. Specific shortcomings in current network
architectures include:
 Too much network hierarchy. Current networks are optimized to support
transactional traffic flows (i.e., traffic between servers and end users) and not
newer application workflows that require connectivity from server to server.
Unfortunately, while this design works well for serving applications and Web
©2011 IDC #232049 3

4. pages to end users, it is poorly designed for the connectivity requirements of
cloud computing and Big Data. Larger, flatter networks optimized for server-to-
server traffic require fewer hops between the servers, which helps reduce latency
and increases efficiency.
 Complexity at the network layer. Many organizations support several different
networks for servers and storage, which can introduce complexity and the need
to support multiple protocols. These could include LAN (Ethernet), storage area
network (Fibre Channel), high-performance computing (InfiniBand), and a
management network (Ethernet). Distinct networks include not just separate
switching products but also separate cabling, network management, and, very
often, staffing. This architecture is necessary due to Ethernet's "best effort"
nature; however, it is costly and can create duplicative costs within the
datacenter and drive up capital and operational costs. One such redundancy is
the duplicative cables that connect a single server to the Ethernet LAN and the
Fibre Channel SAN. As a result, IDC believes that regardless of the ultimate
destination switch, the focus on reducing cost and complexity will manifest as a
migration to Ethernet for both networks: 10GbE is now available with lossless
protocols well suited for storage traffic (both block and file).
 Lack of virtualization awareness. While server virtualization has brought
greater utilization of compute assets, IDC finds that expectations for more
dramatic boosts in operational efficiency often fail to materialize. This is because
the shift to virtualized servers often leads to significant disruptions in a number of
areas, including the overloading of network infrastructure and the
overprovisioning of storage capacity, resulting in a quantum leap in storage
administration costs. Further, conventional networks have lacked the
virtualization awareness required to ensure that VMs can migrate from server to
server, while the unprecedented scale in the number of virtual servers being
supported requires a completely new set of management tools and a rethinking
of the network architecture to accommodate the required traffic patterns. To
create a network architecture that can support these levels of virtualization, the
network must recognize virtual machines and not just physical servers and must
be flexible enough to provide bandwidth, quality of service (QoS), and security to
all virtual machine types without manual intervention.
 Intelligence needed at the network edge. Today, innovations such as cloud
computing and virtualization are happening at the edge of the network, where
physical and virtual infrastructures overlap and where storage and servers
connect to the datacenter network. Restricting network intelligence to the core is
no longer sufficient: With virtual services, intelligence needs to be applied at the
edge to treat traffic appropriately. Innovations at the network edge can drive
improvements in a variety of areas, including network virtualization, management
of virtual machines, and security and network management.
 Complexity in hypervisor options. Today, datacenter managers have multiple
options in hypervisors. While many enterprise customers associate VMware
products (VMware Server, VMware ESX, and VMware ESXi) with server
virtualization, the reality is that many types of hypervisors are available, including
Citrix XenServer; IBM PowerVM; Microsoft Virtual Server, Windows Server 2008
4 #232049 ©2011 IDC

5. Hyper-V, and Hyper-V Server; Parallels Virtuozzo Containers; KVM; and Xen
(open source). To enable IT organizations to choose the virtualization platform
that is most appropriate for the workload, the network must be hypervisor
agnostic, with the flexibility to provide bandwidth, quality of service, and security
to all virtual machine types.
Security Challenges
Security remains one of the top issues that keep CIOs awake at night. As more data
comes online, an increasing amount of sensitive data, such as customer and supplier
information, costs and prices, contracts, and sensitive intellectual property, is at risk.
Organizations are increasingly taking an end-to-end approach to their information
security and as such are requiring security to be an "embedded" characteristic of their
datacenter network infrastructure.
Managing the Whole Environment
As the amount of IT infrastructure in the datacenter increases and as the IT
infrastructure becomes more complex and siloed, managing it becomes a greater
challenge. And yet with the importance of the IT infrastructure to the business, having
the ability to manage it well becomes more important than ever. It is critical to have an
end-to-end application view to understand the status of all resources in the
infrastructure and to be able to pinpoint and mitigate all bottlenecks.
Rigid Network Designs
Agility, the ability to rapidly provision or reconfigure new or existing IT infrastructure
elements, is limited in most enterprise IT infrastructures today. Traditional network
architectures have many layers and are very rigid, making it difficult to rapidly
provision new services. Further, as organizations consolidate their infrastructure
through virtualization, they are currently forced to treat all application traffic similarly.
This has the effect of driving up costs if every workload needs "premium" advanced-
level bandwidth and network services — which defeats the purpose of consolidating
and virtualizing in the first place. The reverse scenario is risky, ignoring the unique
security and policy requirements of mission-critical applications. This can drag down
the infrastructure and severely limit the organization's agility. To meet the agility
requirements and reduce costs, organizations need to maintain the quality of service
for specific application workloads. Additionally, rigid network designs hinder the ability
to scale quickly. IT organizations need the ability to scale the network in equal cost
increments, similar to approaches available in scale-out computing and newer
storage systems. Rigid network designs add additional components and costs to the
network as the network scales.
©2011 IDC #232049 5

6. NEXT-GENERATION NETWORKING MEETING
FUTURE REQUIREMENTS
Modular Approach
Instead of provisioning server, storage, and networking in "silos," a better approach is
to deploy them in "PODs" or "integrated systems" consisting of optimized systems
with each resource working in concert. This approach not only breaks down the silos
of the traditional approach but also allows organizations to scale their IT infrastructure
in a balanced manner (so the amount of compute power scales hand in hand with the
amount of storage and networking resources required). This improves flexibility and
agility, enabling organizations to scale resources up or down as needed.
Flatten Network with a Unified Fabric
Approach
IT organizations must enable the enterprise to respond to changing business
demands quickly and economically, which means the network architecture must
support simpler management and migrations. The best approach to accomplish this is
to flatten and converge the datacenter network and to implement a unified fabric
architecture. This reduces complexity and allows IT administrators to move away from
multiple fabrics, each with separate adapters and cabling (which is expensive and
difficult to manage and reduces network flexibility). Network architects favor flatter
designs to maximize network efficiency, reduce congestion, and address Spanning
Tree limitations by creating active/active Layer 2 network paths for load balancing and
redundancy. IDC believes that the trend toward converging network fabrics on
Ethernet is the right technology path for delivering network-edge technology that can
provide the lossless performance and ultra-low latency needed to meet the demands
of today's mission-critical applications.
Higher Network Speeds
Supporting cloud computing and Big Data requires networks that can handle the
higher speeds needed. As more traffic is introduced into the network, the challenge is
to keep latencies as low as possible as well as increase line rates to keep the network
from becoming the bottleneck.
Figure 2 illustrates the expected growth in higher-speed networking. IDC expects
there to be an explosion in the 10GbE datacenter networking market over the next
five years. IDC expects 10GbE, which represented only 6% of ports shipped in 2009,
to grow to 61% of datacenter network ports shipped by 2015.
6 #232049 ©2011 IDC

7. FIGURE 2
Total 10GbE Datacenter Network Ports Shipped , 2009–2015
18,000
16,000
14,000
12,000
(000)
10,000
8,000
6,000
4,000
2,000
0
2009 2010 2011 2012 2013 2014 2015
Source: IDC's Worldwide Datacenter Network QView, September 2011
Further, 10GbE represents only the next step in an inevitable forward evolution in
networking, with 40GbE just becoming available and 100GbE on the horizon. As
10GbE becomes mainstream at the server access layer, organizations will begin
requiring higher speeds in the core. Organizations concerned with future scalability
and investment protection would do well to prepare for implementing 40GbE and
100GbE technologies in their datacenters. This is particularly true for businesses
looking to drive the lowest latency possible into their systems — for example, financial
trading houses in which millisecond delays in trade times can represent millions of
dollars in lost profits.
While some early implementers are embarking on selective pilot deployments of
40/100GbE switching platforms, IDC expects volumes to pick up in the 2013–2014
time frame as port costs for 40/100GbE deployments become comparable to and
competitive with those for 10GbE deployments.
Support for Server-to-Server Traffic
Most architectures are designed to support traffic between servers and end users in
the more traditional transaction approach, but with new network architectures, cloud
computing, and big data, traffic flows are now more interactive between elements
inside the datacenter. To achieve real-time business-centric analytics, IT workloads
are becoming increasingly interdependent. New network architectures must be
introduced to handle these new types of traffic, which must be integrated throughout
the whole datacenter.
Customers continue to benefit from the tidal wave of innovation around enhancing
Ethernet. As stated earlier, Ethernet is now a viable option for both block and file
storage. Likewise, similar advances have been made to support low latency server-to-
©2011 IDC #232049 7

8. server traffic. Customers are now utilizing RDMA on 10GbE to support high-
performance computing. The combination of low latency and high-throughput
advances in Ethernet has contributed to the technologies leading installations in high-
frequency trading applications in the financial sector.
Support for Virtualization
Despite the widespread proliferation of virtualization in the datacenter, most
organizations' tools to support virtualization are sorely lacking. This point is well
illustrated by considering the primary methods of moving VMs in the datacenter today
(see Figure 3). For most organizations, moving VMs is a largely or entirely manual
process; only 8.4% use automated policy tools, half move them manually with the aid
of tools, and over one in four move them manually without the use of tools.
FIGURE 3
Phase of VM Mobility
Move VMs using
policy tools Don't move VMs
(8.4%) (12.8%)
Move VMs
manually (27.9%)
Move VMs
manually and
using tools
(50.9%)
Source: Three Data Centers – One Vision, IDC #DR2010_T2_MB, March 2010
Organizations require superior approaches to manage their virtualized environments,
with tools that abstract complexity and support VM sprawl management, dynamic
workload scheduling and automation, and capacity planning across physical and
virtual systems. They must support end-to-end visibility across multitier application
environments and performance and root-cause analysis. These tools must be
hypervisor agnostic and enable organizations to migrate network policies along with
VMs as they move from server to server.
By optimizing and automating virtualization and including networking resources in this
automation, IT can drive greater efficiencies throughout the organization. It can break
down the silos and enable the "IT team" to work together more closely, enabling more
8 #232049 ©2011 IDC

9. tasks to be performed by IT generalists and requiring fewer specialized resources to
manage server and storage networks.
Increasing Datacenter Efficiencies
A key trend today is to drive increased efficiency into the datacenter. An important
element of this trend is the move to build "green" datacenters. Introducing intelligence
into the network edge can help reduce overall power consumption, which not only
benefits the environment in a general sense but also reduces operating costs. This is
important to most organizations given the current state of the global economy and is
particularly true in geographies with high-energy costs and strict regulations such as
Europe.
Furthering datacenter efficiencies is the ability to provide unified management across
the fabric of network devices. The devices in the network should be managed,
configured, and provisioned holistically.
Support for Standards
Leveraging a standards-based approach is another important characteristic that
organizations must take into account when architecting their networks. Not only does
support for open standards provide a path for future upgrades and scalability without
locking into proprietary protocols, but choosing a vendor that supports a wide range of
standards opens the door to additional technologies and efficiencies. For example,
support of the emerging OpenFlow standard introduces a greater level of intelligence
in Ethernet networks through a new approach called Software-Defined Networking
(SDN) and can be used to support on-demand "express lanes" for voice and data
traffic that are time sensitive or to combine several fiber-optic links into a larger virtual
pipe for temporarily handling a heavy flow of traffic.
Storage Networking
Organizations thinking about their next-generation networks should not forget about
storage networks. For those that have implemented Fibre Channel storage,
implementing a 16Gb Fibre Channel network can provide an upgrade path for the
future. Fibre Channel–based capacity shipped will growth at a CAGR of 42% from
4.6PB in 2010 to 26.7PB in 2015. However, the explosion in storage on Ethernet is
inevitable. File-based storage is exploding with videos, images, and unstructured
content in the era of cloud computing. This storage will grow to overtake Fibre
Channel and reach 30.7PB in 2015. Adding NAS/iSCSI to the mix and FCoE means
that storage on Ethernet will account for two-thirds of storage capacity in 2015. To
meet the needs of converged data and storage networking, Ethernet networks must
be lossless. Network equipment must support the Data Center Bridging (DCB)
standards to ensure lossless operation. IDC believes that IP SANs using NAS/iSCSI
or FCoE enable customers to standardize on Ethernet and maintain the appropriate
service levels for storage workloads.
©2011 IDC #232049 9

10. DELIVERING THE NEXT-GENERATION
DATACENTER WITH IBM SYSTEM
NETWORKING
To address these challenges, IBM has expanded its System Networking line of
products focused on the datacenter. The IBM System Networking portfolio leverages
IBM's more than 30 years of experience in datacenter networking, management, and
integration, as well as IBM's reputation as a trusted partner for network solutions, its
support of leading management platforms such as Tivoli, and its support for industry
standards.
IBM defines System Networking as the integration of intelligence into compute and
storage systems management. This integration brings the ability to maximize
compute processing power to deliver data analysis in the fastest possible time. The
goal is to make a more general management solution for a customer site, leveraging
technology such as the Tivoli Network Manager, Tivoli Netcool/OMNIbus, and other
components.
Products offered under the System Networking banner include high-performance IBM
System Networking–branded blade and top-of-rack RackSwitch Ethernet switches,
Fibre Channel SAN switches and directors, partner offerings, and networking
software and solutions. The IBM System Networking portfolio was enhanced with
IBM's 2010 acquisition of BLADE Network Technologies, a leader in datacenter
networking.
IBM System Networking solutions are designed to:
 Drive business value. Connecting servers and storage with a high-speed,
intelligent network that is smarter, faster, greener, and easier to manage can
drive improved economics to the business. By improving the flexibility of IT to
more quickly respond to changing business conditions, companies can become
more competitive in the marketplace.
 Improve security. The System Networking line includes a full suite of security
technologies, as well as consulting and managed network security services
designed to ensure network resiliency. This end-to-end, multilayered approach
includes Proventia Network Multi-Function Security unified threat management
software as well as threat analysis and prevention offerings from IBM Security
Services.
 Support an open, standards-based approach. IBM takes an open, standards-
based approach to implementing converged datacenter network designs to
improve interoperability and make it easier to deploy, maintain, and scale
network solutions. It supports virtualization and cloud computing networking
initiatives, including VMready (with IEEE 802.1Qbg – Edge Virtual Bridging),
OpenFlow, TRILL, and DCB.
10 #232049 ©2011 IDC

11. Product highlights include:
 IBM iFlow Director. IBM iFlow Director, announced in June 2010, is a software-
based load balancer and traffic director for the IBM BladeCenter. Offering 10GbE
switching and a 10Gb statistical flow balancer, it is designed to handle
bandwidth-intensive security applications at the network entry point for security
and Web 2.0 appliances.
 IBM virtual NICs (vNICs). Emulex 10GbE Virtual Fabric Adapter II and Emulex
10 GbE Virtual Fabric Adapter Advanced II are the latest elements of the IBM
BladeCenter and System x Virtual Fabric portfolio. These adapters allow
datacenters to reduce the number of switches required inside the BladeCenter
chassis by allowing each physical 10Gbps port to be divided into four virtual
ports. These virtual ports can also be used to provide VM connectivity without
tying up a physical port for each VM. On top of that, the ability of the IBM System
Networking switches to carve up communications back to the Emulex adapter
provides higher availability, increased security, simpler management, and
maximum performance per vNIC.
 Powering IBM Netezza. The synergies extend to the IBM Netezza data
warehousing and analytics solution. IBM System Networking's Ethernet portfolio
is a building block for Netezza, providing a high-speed, low-latency fabric
connectivity infrastructure for high-performance data warehousing and analytics.
CHALLENGES AND OPPOR TUNITIES
Challenges
Like any vendor participating in this fast-moving market, IBM faces a number of
challenges as well as opportunities. IDC believes the top challenge for IBM is proving
how the network helps deliver the appropriate quality of service to virtual traffic.
Customers migrating workloads to a virtualized datacenter environment prefer to
invest in solutions that are shown to be complete and easy to deploy and that have
the full feature set necessary to support a virtualized environment. IBM will need to
demonstrate the completeness of its solution compared with other alternatives
available on the market.
Other challenges include the necessity to demonstrate to customers that their
investment will be sound, not only today but in the years to come. Customers will not
want to invest in technologies that they will be forced to rip and replace in a few years.
To overcome this obstacle, IBM must demonstrate how its offerings are based on
interoperability and open standards and are open to all hypervisors so that customers
understand they will not be stuck with proprietary vendor lock-in down the line.
©2011 IDC #232049 11

12. Opportunities
The key opportunity for IBM is to show how the increasing need to support cloud
computing, business analytics, and high-performance computing requires datacenter
support of dynamic workloads and how these dynamic workloads require a modular,
fabric-based approach to networking. By showing the ability of the System
Networking solutions to prioritize traffic for different applications or storage systems
through virtual pipes, IBM can demonstrate the increased flexibility of the system and
the improved ability to support dynamic, virtualized workloads compared with
previous-generation "fixed not flexible" networking approaches.
This opportunity is magnified when combined with the strength of IBM's core
ecosystem of partners, customers, and OEM partnerships. IBM is one of the top
players in datacenter networking, with customers in some of the largest companies in
the world, including many of the leading banks, trading exchanges,
telecommunications, electronics, and entertainment companies, and has a robust
network of solution provider partners and OEM suppliers.
CONCLUSION
Driven by the needs to support cloud computing, virtualization, and "Big Data,"
organizations are evolving their networks from fixed, data-centric, client/server
topologies to application-driven, dynamic networks. Network architectures are
flattening, bandwidth requirements are increasing, and organizations are building
greater flexibility into their networks, all to increase flexibility to better support rapidly
changing business requirements.
The IBM System Networking portfolio is designed to address these needs using a
modular, standards-based approach. It combines network intelligence with compute
and storage system resources in individual integrated systems, enabling
organizations to scale their network in a balanced, modular manner. IBM's unified
fabric networking infrastructure, which is part of the company's Smarter Computing
initiative, is designed to meet the needs of today's virtualized, cloud-based datacenter
in a scalable, cost-effective manner while providing the agility organizations require to
compete in today's global marketplace.
Copyright Notice
External Publication of IDC Information and Data — Any IDC information that is to be
used in advertising, press releases, or promotional materials requires prior written
approval from the appropriate IDC Vice President or Country Manager. A draft of the
proposed document should accompany any such request. IDC reserves the right to
deny approval of external usage for any reason.
Copyright 2011 IDC. Reproduction without written permission is completely forbidden.
12 #232049 ©2011 IDC
QCW03015-USEN-00