Content in Intel Communities

Intel Continues to Enhance Software Defined Networking (SDN) Capabilities with VXLAN Optimizations in VMware vSphere 5.1

webadmin@intel.com — Tue, 04 Dec 2012 17:40:19 GMT

Now that everyone is talking about Software Defined Networking (SDN) and Network Overlays, I guess it is time to start talking more about what Intel and VMware have already implemented over the past few releases our products. Over the past few years we have been working together to provide the ability to utilize multiple 10GbE connections per host, separate the network control and data plane using the VMware vSphere Distributed Switch, expand the capabilities of the software storage initiators for iSCSI and FCoE, and optimize the use of capabilities enabled by Intel® Virtualization Technologies for Connectivity (Intel® VT-c). I have just released a technology brief covering many of these topics so I won’t repeat them here, but I think one of the new capabilities in VMware vSphere 5.1 needs a little more explanation: the new VXLAN capabilities that allow the extension of networks beyond the current Layer 2 limitations.

Why is it important?

Traditional physical networking technologies can limit the agility and scalability of the cloud and prevent capacity utilization of virtual server systems. Standard VLAN models are also challenging to manage in a high-performance, multi-tenant cloud environment where scalability is necessary and isolation of logical networks is required. Virtualizing the network layer is one way to effectively address these bottlenecks.

One aspect of the VMware Software-Defined Networking (SDN) architecture is the capability to virtualize network components to achieve the agility, scale and performance requirements of virtual cloud environments. It creates an abstracted transport network that is quickly provisioned and easily managed because it is structured by software. These virtualized networks can dynamically scale to demand, attach to specific workloads, and move across virtual environments.

Network virtualization overlay technologies allow compute resources to pool across non-contiguous clusters or pods. You can then segment this pool into logical networks attached to applications and span them across virtual resource pools and physical boundaries—a much more manageable and efficient option than VLANs.

Since these logical networks are decoupled from physical network topology, they can be scaled without reconfiguring the underlying physical hardware. This solves the problem of time-consuming planning cycles for VLAN provisioning. It also helps reduce VLAN sprawl and overcomes the limited number of available VLANs. Operations are simplified and application provisioning can be done more effectively.

To solve these challenges, VMware, Intel and other leading networking and silicon vendors have created the VXLAN technology.

As VMware explains it, Virtual extensible LAN (VXLAN) is a method for “floating” virtual domains on top of a common networking and virtualization infrastructure. By leveraging industry-standard Ethernet technology, large numbers of virtual domains can be created above an existing network, with complete isolation from each other and the underlying network enables administrators to create elastic, logical networks that span physical network boundaries.

VXLAN works by creating Layer 2 logical networks that are encapsulated in standard Layer 3 IP packets. A "Segment ID" in every frame differentiates the VXLAN logical networks from each other without any need for VLAN tags. This allows very large numbers of isolated Layer 2 VXLAN networks to co-exist on a common Layer 3 infrastructure.

Since VXLAN uses a 24-bit identifier, a single network can support up to 16 million LAN IDs. This is much more than the 4,094 limit imposed by VLAN’s specification, providing improved data encapsulation and more efficient network transport and isolation. This encapsulation is done at a VXLAN end-point that currently either resides in the virtual or physical switch.

In the vSphere architecture the encapsulation is performed between the virtual NIC of the guest VM and the logical port on the virtual switch. This makes VXLAN transparent to both the guest VMs and the underlying Layer 3 network. Although it is transparent to the guest and network, the additional encapsulation mechanism introduces certain performance implications.

To address the impact of this additional encapsulation, Intel once again turned to the VMDq capabilities developed by Intel back in 2007 that enabled VMs to receive network traffic at full line-rate on 1Gb and 10Gb ports. VMDq is a hardware assist that is currently used by VMware NetQueue to dynamically offload, based on processor and network load, the routing and filtering of network packets to network controller’s hardware-based receive queues much like RSS with native workloads.

By refactoring the network driver and enabling the function in the hypervisor’s kernel, Intel and VMware have enabled RSS to accelerate VXLAN traffic by distributing the receive traffic among various queues and CPU cores, achieving full 10Gb line-rate. This RSS capability is only available when using Intel® Ethernet Converged Network Adapters on vSphere 5.1. It can be enabled by unloading and loading the module with the vmkload_mod ixgbe RSS=”4” on each 10Gb Intel Ethernet CNA on the server.

In the VMware VXLAN performance brief VMware found that the performance of VXLAN on vSphere 5.1 is very close to a configuration without VXLAN, and vSphere 5.1 with VXLAN configured can meet the demands of today’s network-intensive applications.

Intel also supports VXLAN in our Intel Ethernet Switch FM6000 Series products. As part of our Intel Cloud 2015 on-going cloud data center initiative, with particular focus on automation (virtualization), we can support multiple Network Overlay methods in our Intel® Ethernet Switch FM6000 Series products. These ICs are designed for top-of-rack switches, which are situated at the right place in the network to support these new tunneling protocols. Our Intel® Ethernet Switch FM6000 Series devices contain Intel® FlexPipe™ Technology that can be configured to support new and emerging network protocols, including VXLAN. For more information on the capabilities of the Intel® Ethernet Switch FM6000 family check out Gary Lee’s blog on Wired Ethernet.

I know this only scratches the surface of just one of the new SDN features in VMware vSphere 5.1 so keep an eye out for more white papers and blogs. You can also learn more about what Intel and VMware are doing together:

What’s a Cloud without Full Virtualization using Unified Networking?

webadmin@intel.com — Fri, 18 May 2012 16:20:46 GMT

http://www.youtube.com/embed/oYrlRWMbpO0In a cloud environment, virtualization that stops at the server edge stops short of the potential of cloud computing. To operate a robust cloud environment, you also need to not only virtualize but unify your network and storage connections, so those resources can scale up and down with the needs of dynamic workloads. It’s this fully unified and virtualized combination that gives cloud greater flexibility and agility. This is a key point made in a new Unified Networking video now appearing on YouTube. This lively animation—produced on behalf of Intel—provides a high-level look at the benefits of Unified Networking solutions combining network and storage connections to enable better use of resources and better cloud performance. In animated graphics, the video shows that static network and storage connections can’t reallocate themselves to meet the dynamic needs of cloud workloads, and you end up wasting resources. The video also provides a look at a data center built around Unified Networking over 10 Gigabit Ethernet. The Intel® Ethernet Converged Network Adapter used in this example provides more than just a wider pipe; it’s a smarter pipe. You don’t have to break a 10GbE pipe into smaller static connections when using Intel® Ethernet. It gives you the ability to dynamically reallocate bandwidth to provide the correct ratio of storage and network resources to meet the changing needs of individual workloads. The result? “You never have to over-engineer your data center to power your cloud,” the video tells us. If you can spare 4 minutes to watch this video, you can potentially see a new way to look at Unified Networking and how it can help you avoid inefficient hardware allocation in your cloud environment.

To get further details on this new way of looking at using 10GbE connections that the video outlines, check out the Unified Networking content on our Cloud Usage Models site.

Lowering Blood Pressure in Today’s Data Centers

webadmin@intel.com — Thu, 06 Oct 2011 19:43:00 GMT

In today’s data centers, many organizations maintain an Ethernet network for core networking and a Fibre Channel network for storage traffic. As just about everyone knows, this well-established approach to networking comes with its challenges—in the form of different protocols, different hardware, different management tools, and different skills sets for administration.

In an exploration of the solution to these data center and network challenges, the animated narrator of a new video from Intel expresses the problem in simple terms: “The results are often high complexity, high costs, and high blood pressure.”

So what do you do to lower these dynamics in your data center? This how-to video offers a simple prescription: unified networking based on Intel® Ethernet 10 Gigabit (10GbE) products.

In particular, this video demonstrates how to configure a Fibre Channel over Ethernet (FCoE) storage solution using products from Intel, NetApp, and Cisco. The video uses animated graphics and many screen captures to walk you through the process of creating a unified network based on a tested reference architecture. Along the way, the narrator offers helpful tips to smooth out deployment wrinkles.

If you’re considering moving to a unified network, this video is a great place to begin your planning efforts. In just 17 minutes, you’ll gain an up-close view of how it’s done—and just how easy it can be.

Truthfully, though, we can’t promise that unified networking will lower your blood pressure. But it can sure make life simpler for your network administrators.

For the deeper dive, check out Unified Networking with Intel® Ethernet 10 Gigabit Server Adapters and NetApp Storage. --

It’s now playing on a YouTube screen near you.

Using a standard for I/O Virtualization to deliver 50Gbps of Network Appliance performance

webadmin@intel.com — Tue, 19 Apr 2011 12:05:58 GMT

A little over a year ago, at Intel Developer Forum in Beijing, I worked with Citrix Systems* to demonstrate[1] how the PCI-SIG* Single Root I/O Virtualization and Sharing specification (SR-IOV) standard can overcome I/O limitations seen when virtualizing networking workloads. The goal of this work was to provide virtualized networking platforms that can easily slot into today’s virtualized, multi-tenant data centers but still achieve the network line rates of traditional networking gear. This allows IT to consolidate, but still maintain service levels.

Today, Citrix announced a new appliance platform – NetScaler* SDX[2] – that uses SR-IOV and Intel® Virtualization Technology (Intel® VT) for Directed I/O (Intel® VT-d) to provide a fully virtualized network appliance able to sustain up to 50 Gbps of throughput. That’s full duplex by the way. Like Intel, Citrix measures throughput each way, so others might say sustained throughput of up to100Gbps.

If you’re not familiar with Citrix NetScaler[3], it’s a family of appliances deployed inline within the network to provide load balancing, acceleration, offload and security services. It’s used in front of some of the biggest websites in the world. If you’ve surfed the web, it’s almost guaranteed you’ve used a NetScaler.

For NetScaler, network I/O is the name of game. So, when the NetScaler team set out to virtualize the appliance so that multiple NetScaler instances can be run on the same box, they needed an approach that didn’t compromise the appliance’s throughput. This is where the work Intel and Citrix have done with SR-IOV comes in.

The SR-IOV and Intel^® VT-d support built-in to the Intel^® 82599 10 Gigabit Ethernet Controller used on the Intel® Ethernet Server Adapter X520 series provides for hardware-assisted virtualization. Up to 64 Virtual Functions (VFs) per port are exposed as virtual Ethernet devices directly to the host OS that can be assigned to each NetScaler SDX instance which provides direct I/O access, with the 82599 controller hardware managing device queues and address translation on behalf of each instance. This eliminates the performance overhead and extra management introduced by the need for the device emulation or split-mode driver approaches typically used in virtualization.

The result is that the fully virtualized NetScaler SDX appliances can achieve nearly same throughput as the “bare metal” NetScaler MPX appliances. For example, the NetScaler MPX-21500 is rated at up-to 50 Gbps. Citrix also rates the NetScaler SDX 21500, which uses the exact same hardware platform, at up-to 50 Gbps.

This means that when it comes time to consolidate and virtualize a data center, NetScaler SDX can be used to consolidate L4-7 services. However, using Intel® Ethernet with SR-IOV support, this consolidation doesn’t have to come at the cost of sacrificing instance isolation or performance.

Other SR-IOV Blog Posts

http://communities.intel.com/community/wired/blog/2011/01/19/come-and-get-it-sr-iov-primer-document-updated

http://communities.intel.com/community/wired/blog/2010/09/07/sr-iov-explained

[1] http://www.citrix.com/English/NE/news/news.asp?newsID=1865242

[2] http://www.Citrix.com/netscaler-sdx

[3] http://www.citrix.com/English/ps2/products/product.asp?contentID=21679