by IDC Brad Casemore, sponsored by IBM System Networking
As the use of server virtualization has matured, IT organizations are realizing they cannot achieve an optimized and agile datacenter unless the network architecture is designed to support converged systems. Therefore, any approach to converged systems that leaves networking as a secondary consideration will be doomed to failure. Fortunately, the industry has responded with networking products that are integrated with converged systems architectures. These solutions deliver the high performance required for next-generation datacenters. This approach delivers benefits in several areas, including optimization of physical and virtual resources, reduction in ongoing management of virtual machines, increased security, and IT agility. This white paper describes the business and economic advantages of an intelligent datacenter network for converged system applications.
Complexity has become a roadblock to data center efficiency. In order to have efficient, cost-effective data center operation, IT administrators must integrate and simplify network resources and data center management. The solution to simplifying management, reducing costs, easing complexity and increasing time-to-value lies in addressing all the elements of the system: servers, storage and networking. IBM® Flex System™ provides compute, storage and networking resources in a single environment that is both efficient and easy to manage. IBM PureFlex™ System and IBM PureApplication™ System offer an integrated solution using components from the IBM Flex System portfolio. These components provide advanced networking, storage and virtualization technologies, with flexibility for a variety of workloads.
Data centers are undergoing a major transition toward a smarter, more dynamic infrastructure. More flexible IT architectures are emerging to address the demands of current business applications and new areas, such as cloud computing, multitenancy, bring your own device (BYOD), big data, and analytics. As part of the dynamic infrastructure trend, the role of data center networks is also changing. It is causing businesses to re-evaluate their current networks, which were never designed to handle modern workloads and applications. IBM’s Open Data Center Interoperable Network or ODIN addresses many of the key problems faced by modern data networks, including automation, integration, and management. Compared to classical Ethernet architectures, ODIN fabrics have several distinguishing benefits, which are discussed in this white paper.
During mid-September 2012, Arista Networks and Cisco Systems launched what both claimed to be the industry’s lowest latency switches aimed at big data, cloud, Web 2.0 and high frequency trading markets. Both products are Top of Rack (ToR), 10 and 40GbE switches. Arista’s is based upon merchant silicon, while Cisco utilized custom ASIC. Arista’s 7150 Series claims 10 and 40GbE, 350 nanoseconds Layer 2 or 3 forwarding, VXLAN support and what it calls “flexible forwarding.” Cisco’s Nexus 3548 claims 10GbE, 250ns to 190ns Layer 2 or 3 forwarding, and Cisco Algorithm Boost or Algo Boost technology. At 10GbE speeds, a bit is 1/10 of a nanosecond long, so Arista’s 7150 and Cisco’s 3548 switches delay is 3500 and 2500 or 1900 bits, respectively. That is, these products offer processing delays equal to the time it takes a few thousands of bits to traverse a simple metallic or optical wire at 10Gbs! In this Lippis Report, we review Arista’s and Cisco’s new ToR switches and answer the question: what comes after zero latency switching?
It’s been a few months since VMware acquired Nicira and Cisco launched Cisco ONE. But at the sleepy Interop NY show, IBM and HP expanded their SDN portfolios with the addition of SDN controllers. To date, there are just a few firms with controllers, including VMware, Big Switch Networks, Cisco, HP, IBM, NEC and Nebula. VMware put a value on SDN overlay controllers at $1.26B, which peaked the interest of every venture capitalist as well as network executive; so there’s no surprise to see more controllers entering the market. But what’s occurring is that the controller market is segmenting into OpenFlow and Overlay controllers with little to no awareness and/or interoperability between the two control plains. In this Lippis Report Research Note, we examine the new SDN controllers from Cisco, IBM, Big Switch Networks and HP with an analysis of their evolution.
Create a flattened, converged, virtualized, standards-based network
by IBM System Networking
An open data center with an interoperable network is a flat, converged, virtualized data center network that is based on open industry standards. Instead of under-utilized devices, multi-tier networks, and complex management environments, the modern data center is characterized by highly utilized servers running multiple VMs, flattened, lower latency networks and automated, integrated management tools. New software-defined network approaches (including overlay networks and OpenFlow standards) greatly simplify the implementation of features such as dynamic workload provisioning, load balancing, and redundant paths for high availability and network reconfiguration. Further, high-bandwidth links between virtualized data center resources may extend across multiple data center locations to provide business continuity and backup/recovery of mission-critical data. A highly virtualized data center offers greater resource utilization and lower costs. This new network infrastructure also simplifies management and addresses network issues such as latency, resilience, and multi-tenant support for public and private cloud environments. By taking advantage of IBM’s Open Interoperable Networking (ODIN) design approach, enterprises can design a cost-effective and manageable data center that fully uses the potential of virtualization, and gives customers the flexibility to migrate to federated data centers, in which computing, storage, and net-work resources may be treated as dynamically provisioned resource pools that can be rapidly partitioned into any desired configuration. To learn more about the ODIN, visit: ibm.com/systems/networking/solutions/odin.html
UK analyst firm, Quocirca, describes why fast and intelligent system networks, such as IBM’s, are of critical importance for the smarter data centre. As networking technology has standardised around Ethernet, the perception has grown that it has also commoditised. However, as technical architectures evolve to keep pace with the impact of virtualisation, cloud computing and high-performance workloads, the needs around networking are also changing. This is driving data centre operators to understand how recent performance advances and standards-based innovations in system networking infrastructures can meet the performance, scalability and agility requirements of today’s data centres.
If it wasn’t clear before July 23rd that the Software-Defined Networking (SDN) controller market resides within a cloud stack, then it’s clear now, thanks to VMware’s $1.26 billion acquisition of Nicira Networks. VMware is building a cloud stack that’s first to recognize the importance of virtualized networking and places a very high value on it. The multiple to Nicira’s revenue is, let’s say, very high. Microsoft, IBM, Citrix, Red Hat and the open stack community have just taken notice and are in the process of evaluating Big Switch Networks, Arista Networks, Embrane, Pluribus, Plexxi, et al. In this Lippis Report Research Note, we explore this exciting turning point in the open networking marketplace.
This white paper discusses how Software-Defined Networking (SDN) and OpenFlow are changing the way that networks are being designed, utilized, and managed. According to Clabby Analytics, SDN will radically alter the networking market competitive landscape. Thus, companies need to understand these new technologies to meet rising demand for cloud, big data, analytics and virtualization.
June could not have been a busier month for the networking industry. Cisco launched Cisco ONE, Cloud Connect and Unified Access at CiscoLive. IBM System Networking launched an SDN stack tied into the IBM PureSystems computing platforms of Expert Integrated Systems. And the folks at GigaOm delivered another great Structure conference that brought in cloud thought leaders. In this Lippis Report Research Note, we highlight some of the more significant June of 2012 announcements.
There are a growing number of options to make data center networking more flexible. One option is the use of Edge Virtual Bridging or EVB, which has been standardized in project IEEE 802.1Qbg and championed by IBM, HP, Brocade, QLogic, Emulex and many others. This approach extends the Virtual Ethernet Bridge or VEB sometime called Virtual Ethernet Switch (VES) used by hypervisors to connect VMs to the data center network. Renato Recio, IBM Fellow & System Networking CTO joins me to discuss IBM’s approach to virtual network infrastructure overlays and in particular its support for Edge Virtual Bridging.
The cloud model isn’t about transforming IT. It’s about reinventing the way organizations do business. Organizations in every industry, regardless of size or geography, are embracing cloud computing as a way to reduce the complexity and costs associated with traditional IT approaches. Organizations that approach cloud in a tactical fashion risk adding complexity and inefficiency (not to mention security exposure) due to fragmentation, redundancy and operating silos. Conversely, organizations that embrace cloud strategically—from a business as well as IT perspective—can capture new business value through innovation, flexibility, speed, integrity and security—while reducing cost and complexity.
To deliver the cloud’s full business value, cloud-enabled data centers require speed, flexibility, cost-effective operation and scalability. This paper discusses the technical and business requirements of cloud computing, focusing on the networking layer of the cloud.
The Lippis Report has conducted three open industry test of 10 and 40GbE data center fabric switches at Ixia’s iSimCity. Michael Githens of Ixia interviews Nick Lippis of the Lippis Report to look back on what we have learned after testing eleven products from nine vendors including Alcatel-Lucent, Arista Networks, Brocade, Dell/Force10, Extreme Networks, Hitachi Data Systems, IBM, Juniper Networks and Mellanox/Voltaire. We then look forward as to what the industry will be serving up in 2012 for data center fabrics.
Download “Fall 2011 Open Industry Network Performance And Power Test Report” here.