The Fat Pipe - All Packet Pushers Pods

Quick Takeaways 

12.8 Tbps Line Rate L3 Ethernet Switch ASIC

* tailored to the needs of hyperscalers

* Tomahawk 3 is the world’s first 12.8Tbps switch chip, now sampling to customers, that will accelerate the next

* wave of 100GbE/400GbE hyperscale cloud build-out

* high port radix requires low latency,

* buffer architecture,

* instrumentation,

* 40% less power,

* 75% cheaper cost per 100Gpbs through the use of 50G PAM4 signalling

* Up to 256 x 50G-PAM4/25G-NRZ Dual-Mode Serde

Shrinks an OCP Backpack switch (which is basically 6 x 32 100 front facing switches with an internal spine) into a single switch. So what used 10 ASICs before, is now a single ASIC unit.

Architecture Notes

* architected for hyperscale – performance oriented, latency, L3/label switched fabrics, load balancing.

* But no reqt for programmable data plane.

* driven by Deep Learning / ML / AI

* Hyperscale DC market opportunity -growing

* NVMe over fabrics – the use of SSD in distributed storage means that network performance must be low latency

* Storage advances – NVMe over Ethernet

Compare Trident 3

* Trident 3 – has a programmable pipeline and lower latency but matters less.

* Tomahawk 3 at same power and cost but twice the bandwidth (but no programmability)

* Note that high radix, multi-hop Ethernet is their operation. HPC is building Slim Fly/DragonFly

* 16nm die process, gives an 18 month advantage (?)

Interpreting what this means

* focussing on low latency so its not a flexible architecture for overlays etc needing programmability. You likely won’t see this in the Enterprise (although Cisco could waste money building it into a spine only unit and claim its got magic inside)

* building nearly dedicated ASICs for cloud providers is a reflection of how much they are spending on data centres and the need to build products that are suited to their exact requirements.

* Broadcom has three ASIC families : Tomahawk is for speed and port density. Trident is for programmable features at medium density – ie data centres. Jericho is an ASIC targeting carrier grade needs such as routers.

* See Broadcom is the incumbent, it doesn’t have to be better than the competition. It needs to be good enough so that people don’t switch away (see MS Windows). With Marvell buying Cavium we are seeing consolidation in the ASIC market.

* The ASIC market is a scale up business, the larger volume you can manufacture the better your competitive advantage and that why Broadcom bought Brocade to scale up its Fibrechannel ASIC business and reduce costs.

* I don’t believe that Cisco can compete with Broadcom and the custom cloudscape chips on Nexus (which might be a slightly customised version of Marvell silicon)  doesn’t seem practical anymore when most people are focussed not he software defined parts and not on the hardware. In Enterprise, most people haven’t even gotten to 10G so a 32 port 400G silicon isn’t relevant.

* If you are in the Enterprise, this likely doesn’t matter since most of you don’t need more than 2 switches anyway. Having 2 switches with 32 x 400G ports with limited features but going very very makes no sense because you haven’t scaled out.

Today the Datanauts embark on a weight loss program for the compute stack by getting into unikernels.

Rather than standing up a full operating system, how about packaging your application to run as a binary that can be created in an immutable fashion and updated in a snap? That’s idea behind unikernels.

In particular, we’re looking at one effort called IncludeOS. Our guest is Per Buer, the CEO of IncludeOS. We’ll talk about how unikernels differ from a traditional OS, their benefits and tradeoffs, use cases for the technology, and more. Per also gives an overview of the IncludeOS software.

Show Links:

IncludeOS.org

Unikernel.org

Unikernel Devel.org

Datanauts 063: Unikernels Vs. Containers – Packet Pushers

If you Google for a definition of high performance computing, you’ll find several entries on several sites that can be condensed down to this: high performance computing describes the use of supercomputers or parallel processing algorithms to solve computationally intensive problems.

Today on the Datanauts podcast, we’re going to pick the brains of a person who has worked in high performance computing environments. We’re especially interested in the distributed computing aspects of HPC and the challenges it presents to the infrastructure engineering community.

Our guest is Nick Buraglio, a network architect and engineer at the Energy Sciences Network, the service provider network for the Department of Energy that connects national labs. Nick blogs at The Forwarding Plane.net.

Sponsor: ITProTV

ITProTV brings you the most current IT training with over 2,000 hours of content and more than 125 hours added weekly. ITProTV is the resource to keep you and your IT team’s skills up to date. Visit itpro.tv/datanauts​ ​and use code DATANAUTS30​ ​to get a FREE 7-day trial and 30% off a monthly membership for the lifetime of your active subscription.

Show Links:

High Performance Computing (HPC) – AWS

Best practices for Securing the Science DMZ by Nick Buraglio – YouTube

The Forwarding Plane.net – Nick Buraglio

Let’s get physical, Datanauts. Today we talk about the key elements of building a physical data center: data cabling, racks, HVAC, power, and then managing all of that stuff.

The death of the physical data center is greatly exaggerated, and today we build the perfect data center beast. Our guest architect is John Kerns, Principal Network Engineer for an IT professional services company. John blogs at PackeTsar and co-hosts the InfoTrek podcast. He’s also an active developer on GitHub.

Show Links:

Explain: Tier 1 / Tier 2 / Tier 3 / Tier 4 Data Center – NixCraft

Data Center Power and Cooling White Paper – Cisco

Back to Basics: Power – John Kerns (via Packet Pushers)

Back to Basics: Cooling – Part 1 – John Kerns (via Packet Pushers)

Back to Basics: Cooling  –Part 2 – John Kerns (via Packet Pushers)

Datacenter Vanity – John Kerns (via Packet Pushers)

Today on the Datanauts, we look at non-volatile memory express, or NVMe.

NVMe SSDs are fast. As these drives go mainstream and make their way into enterprises and clouds at scale, what’s that mean for distributed storage networking?

We play “move the bottleneck” on today’s Datanauts. Our guest is J Metz, a.k.a @drjmetz on Twitter. He’s Research & Development Engineer for Advanced Storage at Cisco Systems.

He’s also been on previous episodes, including Datanauts 093: Erasure Coding And Distributed Storage.

We discuss the basics of NVMe and how it works, look at its performance specs, and explore how this technology will impact the network and the design of distributed storage systems.

Today on the Datanauts podcast, we rocket ahead in time. Put on your favorite time-travel helmet, and join us as we explore commodity hardware, NVMe, distributed architectures, and of course, cloud.

Our destination? The future of storage.

Our Guild Navigator for this glimpse into the future is Tom Lyon. He’s the founder and chief scientist at DriveScale and can be found on Twitter.

In part one of the show, Ethan and Chris relay audience questions to Tom, including the long-term prospects of Fibre Channel, what’s keeping Fibre Channel alive, and the implications of distributed storage.

In parts two and three, the Datanauts delve into commodity storage, NVMe, and storage’s cloudy future.

Sponsor: ITProTV

ITProTV brings you the most current IT training with over 2,000 hours of content and more than 125 hours added weekly. ITProTV is the resource to keep you and your IT team’s skills up to date. Visit itpro.tv/datanauts​ ​and use code DATANAUTS30​ ​to get a FREE 7-day trial and 30% off a monthly membership for the lifetime of your active subscription.

Today on the Priority Queue, we’re talking about the book Mastering Python Networking by Eric Chou. Eric also blogs at blog.pythonicneteng.com.

The book aims to help network engineers get more comfortable with automation, coding, and with tools such as Ansible.

Now, lest you think Eric wants you to buy his book so that he can buy a Ferrari, he’s donating all proceeds from the book to charity.

And Eric has a special offer for Packet Pushers listeners. The first 10 subscribers to Eric’s newsletter will get a free ebook version of Mastering Python Networking. And for everyone who subscribes, you’ll be entered in a drawing to win a free autographed copy of the book. Sign up here.

Show Links:

Eric’s Blog

Mastering Python Networking Special Offer

Mastering Python Networking – Amazon

Eric on GitHub

Eric’s Python Repo on GitHub

Mastering Python Networking Book Q&A

Mastering Python Networking on GitHub

Python 2 Death Clock

Today on the Datanauts podcast, our friends at Oracle Ravello fly their sponsorship rocket by our microphones for a discussion of how to stand up a replica of your data center in the cloud.

Why would you want to do that? All sorts of reasons! PoCs, dev/test, user acceptance testing, threat simulation exercises, training, and even to run production workloads in AWS, Google Cloud, and Oracle Cloud without having to make changes to the application.

Our guests from Ravello are Abhi Gupta, Director of Product Management; and Gil Hoffer, VP of Software Development.

We discuss use cases for Oracle Ravello, examine how the product differentiates itself, and drill into the latest product release: Ravello on Oracle Cloud Infrastructure (a.k.a Bare-Metal Cloud).

Show Links:

Oracle Ravello

Sign Up For A Free Trial Of Ravello

Free Ravello Lab For VMware Experts

Follow Ravello on Twitter

We’re always interested in open source projects here at Packet Pushers.

Today on the Priority Queue, we get an overview of ZeroTier, an open source networking project with a commercial variant that claims to “network almost any kind of device or application as if the entire planet is a single cloud region.”

Our guest is ZeroTier founder and CEO Adam Ierymenko. We’ll talk about the origins of the project and how it works. We also explore use cases for home, enterprise, and cloud.

Show Links:

ZeroTier

ZeroTier on Github

ZeroTier on Twitter

How do you provide highly available, scalable services to a cloud-native app that’s been architected with microservices? Just stand up a whole lot of virtual IPs and pools? Not exactly.

You need both more capability and more automation. You need a service mesh. And today on the Datanauts podcast we’re joined by sponsor AVI Networks to find out what a service mesh is and how it supports cloud-native applications.

We’re joined by Ashish Shah, Senior Director of Products at AVI Networks.

Broadly speaking, a service mesh ensures that the various components and services that make up a widely distributed application can reach each other using techniques such as service discovery, load balancing, and handling failures.

The Datanauts and Ashish drill into the major concepts of a service mesh, examine critical components and the architectural approaches, and explore the AVI Networks platform.

Find out more at AVI Networks and check out this special offer for Datanauts listeners: you can claim a free backpack here.

 

Show Links:

What is a Service Mesh and how Istio fits in – IBM developerWorks

What’s A Service Mesh? And Why Do I Need One? – Buoyant

AVI Networks Service Mesh – AVI Networks

Application services for OpenShift-Kubernetes clusters – White paper

Tutorial 1: Deploying Production-Ready Container Apps – AVI Networks

Tutorial Part 2

Tutorial Part 3

AVI Networks on Twitter

AVI Networks on LinkedIn

Ashish Shah on Twitter

Ashish Shah on LinkedIn