Skip navigation

Come and get the latest FPP Whitepaper: Configuring QoS Features with Intel* Flexible Port Partitioning 

 

I went back to see when I published the previous blog on Intel Flexible Port Partitioning (FPP).   It was way back in September!  A lot has happened in this part-time bloggers life since then, including moving on to do things other than virtualization and a trip to Antarctica! 

 

  1. Yes, it’s true, my brother and I went to the bottom of the world; it was a great experience despite the Drake PassageI even have proof, here’s a funny little fellow I took a picture of:

PENG.png

 

It took me a while to get caught up after my month long adventure with penguins, whales, seals and other interesting critters.  I was able to finally finish up the 2nd FPP whitepaper I promised so many months ago – sorry for the delay.

 

This paper entitled Configuring QoS Features with Intel Flexible Port Partitioning  details how to setup various teaming modes, VLANs and rate limiting when using SR-IOV Virtual Functions in your base Operating System or from within a Virtual Machine.

 

I enjoyed writing this paper and I hope that some of you find it of use as you explore using Flexible Port Partitioning using SR-IOV.

 

By way of reminder, I’ve many other blogs regarding SR-IOV, as well as a video explaining the SR-IOV technology and a video showing FPP in action.

Another NEW 10GBASE-T Server Adapter from Intel!


Intel is pleased to announce the availability of another new 10GBASE-T server adapter!  The Intel® Ethernet Converged Network Adapter X540-T1 (single port) joins the recently launched Intel® Ethernet Converged Network Adapter X540-T2 (dual port) as the latest innovation in Intel’s leadership to drive 10 Gigabit Ethernet into the broader server market.  This adapter family hosts Intel’s latest silicon, the Intel® Ethernet Controller X540, which is used by many OEMs as a single chip solution for LAN on Motherboard (LOM) to deliver 10GbE on the latest server platforms.


The MAC+PHY integration drives down both cost and power, enabling broad deployment of 10GbE everywhere in the datacenter.  BASE-T is the form factor that is well understood by the industry and is the easiest and most cost effective to deploy.  10GBASE-T is backward compatible with the customer’s existing network infrastructure, providing a smooth transition and natural migration to 10GbE.


Intel® Ethernet Converged Network Adapter X540-T1 & X540-T2 Key Features:

  • Low pricing - $200 less than Intel’s current solution!
  • 10GBASE-T Single & Dual Port
  • Passive heatsink
  • PCI Express* 2.1 (5GT/s)
  • Standard Cat 6a cabling and higher with RJ45 connectors
  • Energy efficient design – single chip solution with integrated MAC + PHY
  • Unified Networking delivering LAN, iSCSI, and FCoE in one low-cost CNA
  • Flexible I/O virtualization for port partitioning and quality of service (QoS) of up to 64 virtual functions per port
  • Backward compatible with existing 1000BASE-T networks simplifies the transition to 10GbE

 

30% lower power

$200 less expensive

No fan!

 

The new Intel® Ethernet Converged Network Adapter X540-T1 will begin shipping by July 2, 2012.


Backlog orders are being accepted now!

A lot has happened since the Interop* Las Vegas conference in May, where Intel demonstrated its support for software-defined networking (SDN).

 

Now, as we get ready to take the demos to Interop Tokyo, nearly all of the major switching vendors have pledged to support SDN, and International Data Corporation (IDC)* has predicted that the market will grow from $200 million in 2013 to $2 billion in 2016.

 

The primary growth driver, according to IDC Group Vice President Lee Doyle is “highly virtualized network environments and customers who need programmable networks. Customers have always wanted to tune the network, but network management tools have been poor or non-existent."

 

SDN, and its leading protocol OpenFlow, change all of that, giving more management control to multi-vendor networks.

 

At Interop Japan, we will team up with NTT Data* to demonstrate that management control using a combination of NTT Data’s OpenFlow controller and the Intel Barcelona 10GbE top-of-rack switch.

 

The demo will show how Barcelona performs when it is controlled by the NTT Data OpenFlow software control plane.  We’re expecting that Barcelona will deliver its full non-blocking performance on all 48 10GbE ports and four 40GbE ports with latency of only 400ns for L3 switching.

 

Interop Tokyo will run from June 12-15, and I will be there, along with other technical and local experts to answer your questions.

Timing and synchronization have been fundamental components of telecommunication networks for years. Protocols such as SONET and SDH require precise synchronization of all nodes in the network in order to function properly. Ethernet, historically, has not had the need for precise timing because it was built, and predominantly used for non real-time data. 

 

But with the advent of Ethernet cellular backhaul networks, the IEEE* has developed a new precision time protocol (PTP) known as IEEE*1588 V2, to add timing to Ethernet for synchronizing events such as cell tower voice or data handoffs. Now data centers are getting into the timing act.

 

Why do data centers care about precision timing? One class of data center customer is the financial services industry. In financial trading, to use an old phrase, timing is everything. This industry requires that the exact time of trading events can be accurately recorded anywhere in the network down to the microsecond level. But other needs are also arising for precision time measurements in the data center.

 

As I posted in a recent blog, latency is important for applications such as server clustering and certain web services. To guarantee performance for these applications, network administrators can use the new PTP timing protocol to monitor round-trip delays between servers and diagnose network performance problems. This can also help identify points of network congestion and even be used by network management software to rebalance the network load automatically.

 

To service these needs, Arista* just announced it has built a new switch that uses TimeKeeper* software from FSMLabs* to provide PTP solutions to its data center customers. Arista uses our low-latency Intel® Ethernet switch silicon in combination with TimeKeeper to provide the ultimate networking solutions in these emerging datacenter markets.

 

The Arista solution points the way to a data center Ethernet future where low latency and precision timing are combined to help manage application performance.

We’re pleased to announce some innovative new features to our low-latency, Alta-architecture Ethernet switch silicon (the FM6000 series).  Alta based products deliver up to 72 ports of 10 Gbps Ethernet, or up to 18 ports of 40 Gbps Ethernet with latency as low as 400ns. The 24 port FM4000 series solution delivers switching at latencies as low as 200ns.

 

The new functionality includes:

 

Microcode Loader: Ethernet networks are constantly evolving through approval of new standards and functionality.  The microcode loader allows Intel to update Alta switches to support this updated functionality.  Now, new switch ICs have to be spun to support new functionality at wire speed.  But the microcode loader in Alta switches means new functionality can be implemented in the switches in a deterministic manner so there is no impact on performance. The microcode loader is a feature of the Intel® FlexPipe™ packet-processing engine and makes Alta switches ready for ever-evolving data centers.

 

ACL Deep Packet Inspection: With the ability to look deep into a data packet and compare that with access control lists (ACLs), the Alta switch can be used to switch data packets by application, allowing for wire speed differentiated data and traffic classes.  ISPs will be able to use this feature, for example, to offer high priority VoIP or video services.

 

Advanced Hashing:  Popular web applications can have high usage that can swamp a server.  With the advanced hashing – combined with deep packet inspection – these high traffic volumes can be distributed to a number of servers for effective load balancing.

 

The new features are now available for switch developers.  Contact your Intel sales person for more details.

Filter Blog