WHITE PAPER Networking in the Age of Cloud Computing 32 % of organizations surveyed by 451 Research now possess a formal cloud computing plan as part of their overall IT and business strategy. On-premises private cloud adoption accounted for 26 percent of on-premises infrastructure spending in 2013. Hosted private cloud is expected to experience the highest rate of growth for off-premises infrastructure, accounting for 32 percent of hosted spending in the next 24 months. Source: Hosting and Cloud Study 2014: Hosting and Cloud Go Mainstream Introduction 1 Cloud computing, largely enabled by server virtualization, has transformed the IT landscape for enterprises large and small. Long gone are the days when all IT applications were run on dedicated hardware in company data centers. Public cloud services are never likely to meet all IT needs for various reasons, so organizations look to use the flexibility and economy of public cloud services and integrate these with their own private cloud services. These are hybrid cloud services. This scenario has been replaced to a large extent with virtualization: virtual machines are launched and shutdown as required by IT workloads, greatly increasing hardware utilization and enabling organizations to create private cloud computing environments and deploy IT as a service. A virtualized computing environment is in constant flux: virtual machines are brought up and shut down, applications move between physical servers. Until recently data center networks were not similarly fluid, configurations were rigid and could only be changed manually. Virtualization and advances in wide area communication technologies that enable high bandwidth services to be provided at low costs have combined to create the industry that is public cloud. This is because the efficiencies that virtualization offers greatly increase with scale. Now, thanks to the development of Software-Defined Networking (SDN), the network can be configured to optimize links between servers running applications, storage and the external connections to the data center.
Bandwidth Provisioning HOW BANDWIDTH IS PROVISIONED TODAY Today an organization wanting a high bandwidth link between two data centers, whether it s across a city, across a country or across the world, must order the service from a telco, wait a imum of several days for its activation and commit to that link for a imum, in most cases, of 12 months. And that commitment extends beyond bandwidth: QoS parameters like throughput, transmission delay, etc., would also be included. THE NEED FOR FLEXIBLE BANDWIDTH Apart from the demands of cloud computing there are many situations where organizations need bandwidth on demand. Daily back up to a remote location Many organizations choose to back up production data in distant locations for greater resilience. Backups generally take place in batch mode overnight. The link needs sufficient capacity to enable backups to be completed outside of operating hours, but might carry little or imal traffic at other times. With flexible bandwidth, the capacity of the link can be scaled up and down as needed, or it can be disconnected. Disaster recovery For disaster recovery organizations need data, applications and the facilities to support them ready on standby in a location separate from their main facilities. If these facilities are called into service, the organization must also have sufficient communications capacity to make them available wherever they are needed. Product launches New, high-profile online games are highly anticipated by the gag community. Immediately following launch there is massive demand placed on servers as users rush to download the new game. With flexible bandwidth the game developer can respond to varying network demands within utes or even seconds, keeping customers happy and keeping costs under control by not having excess bandwidth when demand has slackened off. Major staff events Large organizations frequently gather executives from widely scattered facilities together into one location for conferences, briefings, etc. All participants will continue to fulfill their normal roles, putting larger than normal loads on the communications networks of the host facility. For example, one global bank has an internship program that gathers 500 interns annually. The event is held in a different location each year, placing a significant additional demand on the connectivity of that facility. Video conferences A company might have a link with bandwidth adequate for nightly backup to a remote site, but if one day it decides it would like to use that link for a video hook-up it could find high latency making this impossible. Short-term processing tasks Organizational demands on IT are not constant: short term projects such as end of financial year processing can place unusually large demands on IT resources where data is spread across multiple locations. Traditional Bandwidth Provisioning Cycle Complex, Manual, Takes Days Customer signs and submits form to order desk PoP-PoP design, build and testing, Racks and power preparation in data center Device installation at customer s rack 30 DAYS Order form Preparation Order validation and acceptance Wiring installation from network platform to customer racks Interconnect testing Figure 1: Traditional bandwidth provisioning 2
THE IDEAL SOLUTION: BANDWIDTH ON DEMAND Ideally long-distance bandwidth services should match the flexibility and fluidity of the organizations they serve and of the IT facilities to which they connect: Links with specified bandwidth, latency and QoS should be able to be established and torn down at will either by human action or under direct control of software. Pricing should be based on the duration of commitment and on committed bandwidth usage but with the ability to increase bandwidth as and when required. A central controller should have overall visibility of and control of the network. All these things are possible with a SDN. SOFTWARE DEFINED NETWORKING: THE ENABLER OF BANDWIDTH ON DEMAND At the heart of all data networks are switches and routers - hardware devices that control the flow of data packets, delivering them to the destinations specified in their address headers. Today most switches and routers contain two major components: the data plane that is responsible for directing the flow of data packets and the control plane that provides the instructions to the data plane. In commercial routing and switching hardware the control plane function is typically fulfilled by a proprietary operating system - for example Cisco s IOS and Juniper Network s Junos. In switches and routers that support SDN the control plane function is removed from the physical device and implemented in software running on standard servers. Standards have been developed for this interface enabling the creation of software that can be used to control a network of routers and switches conforg to the same standard. OPENFLOW: THE ENABLING STANDARD OF SDN To create and maintain an open standard interface between control plane and data plane a group of global major users of communication technology that included Facebook, Google and Microsoft created the Open Networking Foundation (ONF). It has now grown to well over 120 members, including most major switch and router vendors, and has developed the OpenFlow protocol - the first standard communications interface between the control and data planes in a software defined network. OpenFlow: Key to software defined networking OpenFlow is the first standard communications interface defined between the control and forwarding layers of an SDN architecture. OpenFlow allows direct access to and manipulation of the forwarding plane of network devices such as switches and routers.... No other standard protocol does what OpenFlow does, and a protocol like OpenFlow is needed to move network control out of the networking switches to logically centralized control software. Because [OpenFlow] SDN controllers provide complete visibility and control over the network, they can ensure that access control, traffic engineering, quality of service, security, and other policies are enforced consistently across... network infrastructures.... Enterprises and carriers benefit from reduced operational expenses, more dynamic configuration capabilities, fewer errors, and consistent configuration and policy enforcement Source: Open Networking Foundation White Paper: Software Defined Networking the New Norm for Networks There are three main advantages for SDN: It enables applications to directly configure the network to meet their requirements. It provides end-to-end visibility across the network through separating the control plane and data plane. The costs of the routers and switches are greatly reduced because they are standard items that do not incorporate proprietary operating systems to perform control plane functions. 3
The Pacnet Enabled Network (PEN) SDN technology within data centers is rapidly gaining acceptance. However its implementation in wide area networks is in its infancy. The Pacnet Enabled Network (PEN) represents one of the first implementations of SDN platform to commercially available domestic and international wide area network services. Pacnet has deployed OpenFlow-enabled routers and switches in its Asia Pacific network of data centers and has developed software to control and configure these through OpenFlow, to create the PEN. Via a graphical user interface or Pacnet Connect, Pacnet customers can set up links between PEN-enabled data centers throughout Australia, Hong Kong, Japan, Singapore, and now the United States, specifying bandwidths from 1Mbps -10,000Mbps, durations from 1 hour to years and with levels of latency - low, standard or best effort. Pricing is displayed in real time and the requested circuit is activated within utes. Application programg interfaces (APIs) are available to enable the same instructions to be embedded into software. Pacnet Enabled Network (PEN) Simple, Automated, Within Minutes Select Select Switch Build Flow Provision 1 1 5 2 Figure 2: Provisioning can be done in utes Figure 3: An online pricing calculator gives an indication of price for any combination of Pacnet Enabled Network (PEN) route, bandwidth and service quality. 4
PEN AND PACNET AWS DIRECT CONNECT: A POWERFUL INTEGRATION Pacnet's AWS Direct Connect service offers high-performance, secure and cost-efficient access to Amazon Web Services. When combined with the Pacnet Enabled Network (PEN), customers can seamlessly and dynamically match their bandwidth to the demands of bandwidth-intensive workloads and applications hosted in Pacnet s data centers connected to AWS cloud. AWS Direct Connect gives customers hosted at Pacnet s Tier III data center facilities in Singapore, Sydney and Tokyo, as well those connected to any Pacnet Points-of-Presence (PoPs) across the Asia-Pacific region via Pacnet's Global Ethernet International Private Leased Line (EIPL) and MPLS-based Global IP VPN services, the ability to dynamically manage application workloads across the AWS cloud and their private IT environments. By using a dedicated private network to access AWS, Pacnet AWS Direct Connect delivers a more consistent network experience and improved performance and lower overall network costs because customers save on public Internet connectivity. Case Study Online retailer uses PEN with AWS Direct Connect This Pacnet customer, a leading kitchen and laundry specialist founded in 1906, has been running an online store since 2005. It has made a big investment in Agile development and deployment methodologies and is a heavy user of Amazon Web Services: Amazon S3 (public cloud storage) and AWS Virtual Private Cloud. These services are accessed using AWS Direct Connect with 200Mbps of bandwidth, providing 100Mbps for its main Virtual Private Cloud (VPC), 50Mbps for its secondary VPC and 50Mbps for access to public assets like S3. The company connects branch offices and retail stores over a Pacnet MPLS network. This network provides access to the company's billing and order management systems as well as important digital assets for the appliances they sell, including (photos, manuals, etc - stored in AWS S3). Now, the company is using PEN to provide connectivity between AWS Direct Connect and its MPLS network. By connecting its 200Mbps AWS Direct Connect into a 100Mbps PEN Long Duration Flow service, the company can grow the PEN flow on demand without paying for bandwidth it does not need. 5
WHY AUSTRALIA NEEDS ELASTIC BANDWIDTH Organizations in Australia stand to gain more than most from the flexibility in bandwidth provisioning provided by the Pacnet Enabled Network, thanks to the tyranny of distance. Most large organizations have facilities in all the major cities of the countries in which they operate. Facilities in one city might act as a backup and disaster recovery site for those in another, resources might be replicated for better performance, very large files (engineering designs, graphics, etc depending on the nature of the business) might need to be transferred between locations. All these applications create huge peaks and troughs in bandwidth utilization. All Australia s major cities are hundreds of kilometers apart, bandwidth costs are high and therefore inefficiencies resulting from bandwidth capacity that is under-utilized are great. For organizations operating internationally, the distances between Australia s major cities and those of important trading partners are also much greater than for nations in the Northern Hemisphere, and with Australia and New Zealand the only sources of traffic on these routes, costs are further increased above those for comparable distances on much more heavily used routes in the Northern Hemisphere. The nearest capital city to Australia, Auckland is almost 2,000kms from Sydney. Going northwards, Indonesia s capital Jakarta is 5,200kms from Melbourne. Beijing - capital of Australia s largest trading partner, China - is more than 9000kms from Melbourne. Tokyo - capital of Australia's number two trading partner - is more than 8,200kms from Melbourne. Case Study Techno Group The Techno Group, one of Australia s leading systems integrators, is using PEN for its disaster recovery operation. It offers private cloud services hosted out of data centers in Melbourne and Sydney for clients throughout Australia and the wider Asia-Pacific region. Prior to taking up PEN Techno Group used a permanent connection for its disaster replication requirements. This link carried traffic only during disaster recovery, but incurred the same fixed bandwidth cost, 24 hours every day. With PEN Techno Group is now able to provision network services between two locations in utes via the PEN online portal. The Techno Group is also using PEN to support design and printing companies that need to send large files across continents. The Techno Group is only billed according to the amount, quality and duration of its network usage while at the same time benefitting from Pacnet's industry-leading SLAs and secure private network. The Pacnet Enabled Network has taken our disaster recovery program to a new level by giving us the flexibility to perform data replication based on demand. Peter Bell Founder Techno Group 6
Summary Virtualization has greatly increased the efficiency and flexibility of computing resources. Instead of processors and storage being dedicated to specific applications they can be shared among many applications. To further increase the flexibility of virtualized data centers the networks within them that connect servers to storage and to the outside world can now be controlled by software to meet the changing needs of the applications they serve. The growing popularity of cloud computing, the need for remote backup facilities, disaster recovery and the increasing fluidity and mobility of global business all combine to create rapidly changing and often unpredictable demands for bandwidth between data centers in different cities and countries. Until recently such flexibility was not available: bandwidth had to be acquired on long term contracts, provisioned manually and took days to activate. PEN: REVOLUTIONIZING BANDWIDTH PROVISIONING The PEN underpinned by SDN technologies has changed that and has brought about a revolution in long distance bandwidth provisioning. Under user control from a simple and easy to use graphical user interface - or directly by software - links can be provisioned in utes; for an hour, a week, a day, up to years, and bandwidth and QoS parameters changed on the fly. By removing the need to have dedicated bandwidth locked in by long term contracts and lying idle PEN can bring significant savings to corporate communications budgets. By linking bandwidths costs directly to bandwidth usage requirements PEN can make affordable services that were previous prohibitively expensive and open up possibilities for new bandwidth-dependent commercial offerings. The possibilities are endless. Contact Pacnet today to discuss how PEN can expand your horizons. ABOUT PACNET Pacnet is Asia-Pacific s leading provider of managed data connectivity solutions to cloud and content service providers, multinational enterprises and telecommunications carriers. Ownership of the region s most extensive high-capacity submarine cable systems with over 46,000 km of fiber and interconnected data centers across the Asia-Pacific region gives Pacnet unparalleled reach to major business centers in key markets including China, India, Japan and the United States. Combined with a comprehensive set of managed network and value-added data center services, its assets and experience in the region help Pacnet service large businesses worldwide including Fortune 500 companies. Pacnet is headquartered in Hong Kong and Singapore, with offices in all key markets in the Asia-Pacific region and North America. CONTACT US For more information, contact your Pacnet Representative or email us at sales.inquiries@pacnet.com CONNECT WITH US v100 2014.10.01 Dtm www.pacnet.com