The document discusses the shift towards more design-driven and innovative data center solutions as enterprises turn to cloud computing to accelerate business innovation and contain costs. It notes that data from a growing number of sources is pouring into data centers, requiring more flexible architectures to handle diverse and real-time workloads. It outlines some of the key challenges with traditional hierarchical data center designs in supporting modern requirements around server-to-server communications, performance, security and availability. Finally, it discusses emerging trends like data center consolidation, virtualization, new application models, and the need for low-latency, high-performance networks optimized for the cloud.
Data Center Solutions: Radical Shift toward Design-Driven Innovation
1. Data Center Solutions: Radical Shift toward Design-Driven Innovation
Friday, 14 September 2012
Enterprises are turning to the cloud to accelerate business innovation, improve business
agility, and contain costs.
Leveraging data in innovative ways presents unparalleled opportunities to understand and
respond to the needs of a rapidly moving and changing market. As an increasing volume of
data from an ever-growing number of sources is pouring into data centers, it is no longer
efficient for data warehouses to manage single, homogeneous workloads. Business analysts
need access to a wide variety of data, in real time, for diverse uses such as standard and
specialized reporting, online analytical processing (OLAP), advanced analytics such as data
mining and data marts, and staging for detailed or real-time data sources. The strategic
requirement for real-time data warehousing also adds online transaction processing (OLTP)
workloads to the mix, furthering the strain on existing infrastructures.
Cisco, Dell, Delta, Eaton, Emerson, EMC, Extreme Networks, HCL Infosystems, HP, IBM,
Microsoft, R&M, Rittal, Schneider Electric, TE Connectivity, and Wipro Infotech are a few
data center solution providers in India. The major data center service providers are
Reliance Communications, Tata Communications, Sify Technologies, Netmagic Solutions,
Airtel Business, Tulip Telecom, CtrlS, TriMAX, and SIS.
Business and Technology Drivers for New Data Center Architectures
With so much data being used in so many ways, the IT infrastructure has to accommodate
the constantly fluctuating demands on processing and storage. Traditional distributed
approaches to infrastructure and architecture lack flexibility required to maintain
processing speed and scalability. As a result, IT organizations end up spending precious
time, building up resources, and reallocating excess capacity to support shifting
workloads.
Data center consolidation. With mission-critical applications and services deployed to
provide the foundation for day-to-day operations and delivery of end-customer services,
the data center needs to deliver exceptional availability and meet stringent service-level
agreements. Exploiting server virtualization and low-cost computing power, customers are
deploying more and more sophisticated applications on a larger scale. To reduce the sheer
complexity and improve operations of these deployments, customers are seeking to
consolidate fragmented, dispersed facilities into fewer, centralized locations. These new
mega data centers are fundamentally challenging how networks must be built. Networks
must be designed to deliver much higher levels of performance, scalability, and
availability than previously expected in order to meet service-level agreements (SLAs) and
maintain continuity of operations. Beyond sheer performance, these data center networks
must quickly recover from hardware- or software-related faults and protect against server,
storage, network, and application vulnerabilities to help ensure continued performance
and minimized service disruptions.
Increasingly powerful and scalable compute. The adoption of increasingly powerful
multicore processor servers, higher-bandwidth interfaces, and blade servers is
dramatically increasing the scale of data center deployments. Now, thousands of virtual
machines can be deployed in a single data center to consolidate infrastructure and
streamline operations. These large-scale solutions are dramatically increasing network
2. performance requirements at the server edge and across the extended network. Likewise,
virtualization and vMotion/Live Migration tools for moving virtual servers are introducing
high-volume machine-to-machine traffic flows and impacting existing administrative
practices, creating a new virtual edge that blurs the traditional boundaries between
network and server administration.
New application deployment and delivery models. Traditional client-server software and
infrastructure deployment models are being displaced by new application architectures
and service-delivery models that are reshaping the data center. Web 2.0 mashups, SOA
solutions, and other federated applications are being widely deployed to deliver
integrated, content-correlated, context-specific information and services to end-users
within the enterprise and beyond. These deployments drive new, bandwidth-intensive
traffic flows within the data center and demand low-latency, high-performance server-to-
server and intra-server, virtual machine-to-virtual machine connections. At the same time,
cloud computing and XaaS initiatives are introducing more stringent service-level security
demands and driving requirements for a more agile and dynamic infrastructure.
Current Legacy Data Center Design Challenges
Today's data center networks were designed to support conventional siloed IT
architectures in which servers are dedicated to specific functions or organizations and the
vast majority of traffic flows in and out of the data center. Most data center networks are
based on hierarchical designs comprised of an access tier, an aggregation tier, and a core
tier.
The access tier is made up of Ethernet switches connecting rack servers and IP-based
storage devices (typically 10/100 Mbps or 1GbE connections). The access switches are
connected via Ethernet to a set of aggregation switches (typically 1/10GbE connections),
which in turn are connected to a layer of core switches or routers that forward traffic to
an intranet, the Internet, and between aggregation switches. Layer 2 VLANs are typically
implemented across the access tier and aggregation tier, and Layer 3 routing is
implemented in the core.
Bandwidth is typically over-provisioned in the access tier and, to a lesser extent, in the
aggregation tier. The server and the networking infrastructure are typically administered
independently, by separate teams using distinct toolsets. Each server is dedicated to a
specific function (e.g., Web server, application server, database server) and can be
reasonably well protected using conventional security solutions such as intrusion
prevention systems. Hierarchical networks are not well suited for server-to-server
communications. Server-to-server traffic must traverse multiple layers of switches.
Contemporary data center networks designed to support simply siloed IT architectures can
not meet the performance, security, availability, and agility requirements of the cloud.
Public cloud service providers and enterprises deploying private clouds must implement
simpler and more efficient networks that support the bandwidth-intensive, delay-sensitive
server-to-server traffic flows and stringent SLA and security demands that accompany
cloud computing. They must also adopt new management systems and practices to
orchestrate on-demand services and administer and isolate virtual resources.
Low-latency server-to-server connections; greater performance and resiliency; higher
bandwidth at the server edge; reliable application delivery across public or private cloud;
high availability, business continuity, and disaster recover; unified management, and
virtualization-aware security are the few requirements of the cloud-optimized data center
3. network.
Low-latency server-to-server connections. The three-tier hierarchical networks are not
well suited for high-volume server-to-server communications. Inter-server traffic is forced
to traverse multiple layers of switches, and each switch adds latency to the connection.
Enterprises and service providers need to implement flat, low-latency data center
networks to accommodate the delay-sensitive, volume-intensive server-to-server traffic
flows that accompany cloud computing models.
Greater performance and resiliency. Hierarchical data center networks typically rely on
some variant of the spanning tree protocol (STP) for resiliency. STP is designed to allow
only one active path from one switch to another, regardless of how many actual
connections might exist in the network. If the active path fails, the protocol automatically
selects a backup path. STP can take several seconds to recover from link failures and is not
well suited for delay-sensitive applications. Cloud computing requires more efficient and
resilient network designs that make full use of networking resources (no idle backup paths)
and recover from failures in milliseconds to meet stringent resiliency requirements.
Higher bandwidth at the server edge. Blade servers and server virtualization pack more
and more computational power into smaller and smaller form factors like - increasing
bandwidth demands at the server edge, driving requirements for new switching solutions
that offer better performance, and greater port densities.
Reliable application delivery across public or private cloud. Organizations responsible for
private cloud services must deliver predictable and reliable network services to support
mission-critical business applications and address the objectives of internal service-level
commitments.
High availability, business continuity/disaster recovery. With hundreds or even thousands
of virtualized applications now in play across multiple, consolidated data centers, network
resiliency and high availability take on a new, heightened level of importance. Network
platforms and designs need to recover quickly from hardware and software faults to
maintain service and business continuity. Enabling distributed workloads and replicating
data and applications across multiple, geographically dispersed data centers are a
challenge.
Converged network infrastructure. With the proliferation of virtualization and the
escalating demand for communication and storage performance, I/O sprawl has IT at the
breaking point. The traditional model of completely parallel, autonomous data and storage
networks with dedicated interface cards, switches, and cabling plants can be costly and
inefficient. Enterprises are looking to consolidate server and storage connectivity to
reduce equipment and operating expenses; eliminate clutter and complexity; and make
efficient use of shared networking resources while ensuring continuity of service.
Unified management. Server virtualization creates a new virtual edge that blurs the
traditional boundaries between network and server administration and introduces a variety
of operational challenges. New tools are required for efficiently administering virtual
switches, servers, and connections; for orchestrating on-demand applications and services;
and for ensuring SLAs and enforcing service policies as VMs migrate across the data center.
Virtualization-aware security. The virtual edge is beyond the scope of existing security
systems and practices. In contemporary data centers, distinct workloads and tenants are
deployed on discrete physical servers. Workload-to-workload communications always occur
over physical connections and can be secured using conventional intrusion prevention
4. tools. With server virtualization, workloads can communicate over virtual connections
within the same server in a manner transparent to existing network-based intrusion
prevention systems. New virtualization-aware security solutions to police intra-server
communications flows, protect virtual resources, and partition multitenant environments
must be considered.
I/O Consolidation and Network Convergence
With the proliferation of virtualization and the escalating demand for communication and
storage performance, I/O sprawl has IT at the breaking point. The traditional model of
completely parallel, autonomous data and storage networks with dedicated interface
cards, switches, and cabling plants can be costly and inefficient. Enterprises are looking to
consolidate server and storage connectivity to reduce equipment and operating expenses,
decrease clutter and complexity, and make more efficient use of shared networking
resources while ensuring continuity of service.
In addition to having the potential to reduce complexity at the network edge, FCoE also
promises to reduce equipment costs beyond the access layer by enabling convergence in
the extended network. However, given the critical nature of storage networking and its
central importance to the integrity of the end-to-end data center architecture, customers
should carefully evaluate the implications of these solutions when considering more
extensive FCoE deployments.
Enterprises are turning to the cloud to accelerate business innovation, improve business
agility, and contain costs. Cloud computing reshapes the way applications are deployed
and consumed, and influences data center network designs. Radical shifts in thinking on
data center design and sourcing are under way, in part spawned by cloud providers at the
extreme end of data center design and scale. Fundamentally, new data center designs are
all focused on energy efficiency across power, space, resource utilization, cooling, and
capital expenditures.
EXPERTS SPEAK
Intelligent Data Centers
With the need for IT services permeating itself into all sectors of
the economy, remote and continuous access to real-time information is the key to driving
critical businesses. The cloud has brought about business changes with an unprecedented
5. need to rapidly and reliably scale up capacity. Uncertainty has been intricately weaved
into the high growth nature of the business; hence, the need for flexible data center
infrastructure that guarantees almost cent-percent uptime. A case in point is the
proliferation of e-commerce websites, which is strongly linked to the performance of the
back-end IT infrastructure for their success.
As the business requirements of the users shift and expand workloads within the data
center, they need the ability to move critical resources such as power or computing to
match server demand rapidly. Data center infrastructure management (DCIM) is probably
the most critical element of the strategy to cope with the rate of change and the
complexities that come with it, as well as to minimize power usage effectiveness (PUE)
and reduce the operational cost. Integrated DCIM tools provide managers real-time insights
into their IT and facility infrastructure, as well as monitor and control every piece of
critical infrastructure in the data center with respect to critical resources. That level of
insight gives businesses the extra bang for every watt of power consumed in the data
center.
Dynamic DCIM can optimize all the three layers - infrastructure, physical, and IT in the
data center. This ensures real-time monitoring on capacity, usage, asset, inventory,
configuration, and energy thereby resulting in less power waste, low rates of down-time
and longevity of the infrastructure. At the same time, periodic assessment of the current
infrastructure with respect to industry standards and best practices is very important.
Deploying smarter technologies such as scalable and capacity modular cooling
infrastructure or intelligent power solutions that can talk to other pieces of the physical
infrastructure are also crucial steps in building data centers of tomorrow.
Current business trends point toward more customers opting for design, build, and turnkey
approach with SLA commitments. Predictive monitoring tools are going to become the
norm in every data center. There is also a healthy adoption of modular intelligent and
integrated infrastructure. Vendors are demanded to commit PUEs at different capacities
while designing and building the data centers. Customers are building data centers to
meet global tier standards so as to ensure committed uptime and availability of
infrastructure.
While achieving a low PUE is going to be the ultimate aim of tomorrow's data centers, PUE
does not capture how reliable the data center is or how long the servers will last. With the
wrong site selection, lack of full proof design, implementation, and adoption of best
practices, efficiency measures such as PUEs are nothing more than technical jargons
designed to encourage clients into green facilities based on economic sense.
EXPERTS SPEAK
Rising Domestic Requirements - The Primary Growth Factor
6. Running data centers for any enterprise involve processes, proper
technologies, and a dedicated team to look after it; thus, it becomes difficult to manage
cost for companies that are managing it in-house. According to an IDC report in 2011, third
party data centerĀ¬ā services are expected to reach USD 671 million by the end of 2012,
which is growing annually by 37 percent. In 2009, third party data center space stood at 2
million sq. ft. and was estimated to grow at a CAGR of over 30 percent to reach around 9
million sq. ft. by 2014. Looking at the current trend, a lot of enterprises are expected to
move toward data center outsourcing to optimize cost. Realizing the huge market
potential, the company embarked on a mega-project to redefine the data center industry
in India by making Tulip Data City. Furthermore, the enormous scale of the facility makes
it the most sought-after solution for virtualization and consolidation needs for majority of
corporate houses in India, given in a scenario when less than five percent of data centers
in the country are present in a professional data center environment.
On partnership and alliances
The formation of strategic alliances or partnership by data center providers has been seen
as a response to globalization and increasing uncertainty and complexity in the business
environment so that each benefits from the strengths of the other and gains competitive
advantages. Enterprises that enter into a strategic alliance usually expect to benefit in
one or more ways. Some of the benefits that enterprises could achieve include achieving
synergy, sharing the financial risk, gaining capabilities, ease of operations, and the agile
infrastructure.
On emerging trends
In the current scenario, data center growth is primarily driven by increasing domestic
requirements from sectors such as financial institutions, telecom operators,
manufacturing, and services. Companies are also investing in additional data centers to
enhance or meet disaster recovery and business continuity requirements. The key growth
driver for the data center market is the ongoing investment in large captive data centers
coupled with the capacity growth witnessed within the data center service provider space.
It has been observed that there has been a significant need for increased data center
efficiency, cost containment, adoption of cloud computing, and SaaS. These factors are
changing the landscape of data centers, IT organizations and their budgeting, and
infrastructure, thus creating enormous opportunities for both cloud infrastructure and
SaaS vendors. The benefits of virtualization include better utilization of server hardware,
reduced floor space requirements, lower power and cooling costs, and improved
7. productivity of data center personnel.
EXPERTS SPEAK
Digital Content to Increase by 50 Percent in 2012
The pace of business is accelerating rapidly. Globalization and
knowledge-based economies are forcing companies to embrace new business models and
emerging technologies to stay competitive. As per a recent data center study by IBM,
2.7ZB of the digital content will exist in 2012, a 50 percent increase from 2011. Clearly,
data is continuing to grow putting pressure on infrastructures and demands on the
organization to turn this data into insights.
Most IT infrastructures were not built to support explosive growth in computing capacity
and information that are seen today. A large number of data centers have become highly
distributed and somewhat fragmented. As a result, they are limited in their ability to
change quickly and support the integration of new types of technologies or to easily scale
to power the business as needed.
With the increased demand for IT, more than 50 percent of mid-size and large-size
companies are looking to expand their data center capacity in the next 12-24 months.
Optimizing the IT infrastructure is the key to re-architecture data centers in most of the
enterprises. Building or upgrading a new data center provides an opportunity to rationalize
the data center strategy perfectly as a way to gain major capital and operational savings,
including energy-efficiency savings. Organizations have realized that they can lower the
total data center cost by 30 percent with better data center designs. This has led to
enterprises looking at re-designing their data centers.
On emerging trends
While data centers across enterprises have seen increased server deployment, its compute
utilization performance is still below 15 percent average but requires the same amount of
power and cooling as a server operating at 75 percent utilization. It calls for server
consolidation.
The trends for Web2.0 and cloud computing is clear and growing and it will need server
8. virtualization as the foundation layer. It will also add to complexity at the system
management layer in a heterogeneous environment. The trend for analytics is seeing
growth and it will bring in need for high-performance computing. It will also bring in
insight into new trends faster and on real-time basis. This will eventually result in IT
supporting innovation in business.
EXPERTS SPEAK
Technology and Automation Mandatory for Efficiency
There has been a growing trend of enterprises choosing to outsource
the collocation and management of their IT infrastructure to third-party data center
service providers. As per IDC, the third-party data center market is scheduled to grow at
36.5 percent y-o-y as compared to the captive data enter market that is expected to grow
at only 18.1 percent y-o-y. The factors that will continue to fuel this growth are - data
explosion, compliance needs, drive within enterprises to improve customer service through
deployment of IT, increasing power, real-estate and human resources cost, demand for
BCP and DR, and falling bandwidth prices.
On key trends
Consumerization of IT, increased focus on green initiatives, continued virtualization,
increased adoption of cloud, and acceleration in converged infrastructure are some of the
key trends expected to impact the data center market.
The new crop of input devices like tablets and smart phones will force data centers to
start investing heavily in technology to provide the needed integration and security.
Data centers will look for ways to not only control costs but also to promote themselves as
green; it has been estimated that energy accounts for 12 percent of data,center expenses.
The number of new virtual servers being deployed probably exceeded the number of new
physical servers deployed last year and the gap will only grow.
Cloud services comprising SaaS, PaaS, and IaaS are expected to grow from USD 110 million
(Iaas and PaaS form USD 44 million) to USD 1 billion (PaaS and IaaS will be to USD 440
million). The impact will depend on what cloud model is adopted (private, public, or
hybrid). Data centers will have to use more of technology and automation to improve their
efficiency.