Reimagining ‘Show IP Interface Brief’
Welcome back! In my first post here on Packet Pushers (Applying A Software Design Pattern To Network Automation – Packet Pushers) we explored the Model View Controller (MVC) software design pattern and how it can be applied to network automation. This post will go a little deeper into how this is achieved and the mix […]
The post Reimagining ‘Show IP Interface Brief’ appeared first on Packet Pushers.
Fine tuning a T5 text-classification model on colab
One of the most interesting recent developments in natural language processing is the T5 family of language models. These are transformer based sequence-to-sequence models trained on multiple different tasks. The main insight is that different tasks provide a broader context for tokens and help the model statistically approximate the natural language context of words.
Tasks can be specified by providing an input context, often with a “command” prefix and then predicting an output sequence. This google AI blog article contains an excellent description of how multiple tasks are specified.
Naturally, these models are attractive to anyone working on NLP. The drawback, however, is that they are large and require very significant computational resources, beyond the resources of most developers desktop machines. Fortunately, google makes their Colaboratory platform free to use (with some resource limitations). It is the ideal platform to experiment with this family of models in order to see how their perform on a specific application.
As I decided to do so, for a text-classification problem, I had to pierce together information from a few different sources and try a few different approaches. I decided to put together a mini tutorial of how to fine-tune a T5 model Continue reading
Fine tuning a T5 text-classification model on colab
One of the most interesting recent developments in natural language processing is the T5 family of language models. These are transformer based sequence-to-sequence models trained on multiple different tasks. The main insight is that different tasks provide a broader context for tokens and help the model statistically approximate the natural language context of words.
Tasks can be specified by providing an input context, often with a “command” prefix and then predicting an output sequence. This google AI blog article contains an excellent description of how multiple tasks are specified.
Naturally, these models are attractive to anyone working on NLP. The drawback, however, is that they are large and require very significant computational resources, beyond the resources of most developers desktop machines. Fortunately, google makes their Colaboratory platform free to use (with some resource limitations). It is the ideal platform to experiment with this family of models in order to see how their perform on a specific application.
As I decided to do so, for a text-classification problem, I had to pierce together information from a few different sources and try a few different approaches. I decided to put together a mini tutorial of how to fine-tune a T5 model Continue reading
Circular Dependencies Considered Harmful
A while ago my friend Nicola Modena sent me another intriguing curveball:
Imagine a CTO who has invested millions in a super-secure data center and wants to consolidate all compute workloads. If you were asked to run a BGP Route Reflector as a VM in that environment, and would like to bring OSPF or ISIS to that box to enable BGP ORR, would you use a GRE tunnel to avoid a dedicated VLAN or boring other hosts with routing protocol hello messages?
While there might be good reasons for doing that, my first knee-jerk reaction was:
Circular Dependencies Considered Harmful
A while ago, my friend Nicola Modena sent me another intriguing curveball:
Imagine a CTO who has invested millions in a super-secure data center and wants to consolidate all compute workloads. If you were asked to run a BGP Route Reflector as a VM in that environment, and would like to bring OSPF or ISIS to that box to enable BGP ORR, would you use a GRE tunnel to avoid a dedicated VLAN or boring other hosts with routing protocol hello messages?
While there might be good reasons for doing that, my first knee-jerk reaction was:
Fifty Years On
What's likely to happen in computer networking in the next 50 years? Lets polish up the crystal ball and see what awaits!Hedge 105: Johan Gustawsson and Changing Provider Architectures
Many service providers have the feeling that they “didn’t do anything wrong, but somehow we still lost.” How are providers reacting to the massive changes in the networking field, and how are they trying to regain their footing so they can move into the coming decades better positioned to compete? Join Johan Gustawsson, Tom Ammon, and Russ White as we discuss the impact of merchant silicon and changing applications on the architecture of service providers.
Traffic Sequence: Which Product Runs First?
“Which came first, the chicken or the egg?” It’s one of life's great questions. There are hundreds of articles published which conclude with eggs predating chickens by millions of years. Unfortunately, Cloudflare users don't have New Scientist on hand to answer similar questions.
Which runs first, Firewall Rules or Workers? Page Rules or Transform Rules? Whilst not as philosophically challenging, the answers to these questions are key to setting up your Cloudflare zone correctly. Answering them has become increasingly difficult as more and more functionality is added, thanks to our incredible rate of shipping products. What was once a relatively easy to understand traffic flow exploded in complexity with the introduction of products such as Workers, Load Balancing Rules and Transform Rules. And this big bang of product announcements is only accelerating each year.
To begin addressing this problem, we developed Traffic Sequence. Traffic Sequence is a simple dashboard illustration which shows a default, high-level overview of how Cloudflare products interact. Think of this as your atlas, rather than your black cab driver’s “Knowledge”. This helps you understand that London is in the south east of the UK, but not that it's quicker to walk than use Continue reading
Automating Data Center VXLAN/EVPN Using CI/CD: Gluware LiveStream Video [6/8]
Chris DiPaola, Senior Systems Engineer – Network at Acuity, chats with Ethan Banks of the Packet Pushers about Acuity’s EVPN/VXLAN network. Chris & his team used the Gluware API to automate their EVPN deployments, all while tied into their company’s CI/CD pipeline. If Gluware might be a fit for your network automation needs, visit here. […]
The post Automating Data Center VXLAN/EVPN Using CI/CD: Gluware LiveStream Video [6/8] appeared first on Packet Pushers.
Day Two Cloud 120: Web Assembly, K8s Rivals, And Other Cloud Computing Trends
On today's Day Two Cloud we talk trends and predictions in cloud computing, including emerging technologies such as Web assembly, rivals to Kubernetes, and the role of GitOps in infrastructure as code. Our guest is Adrian Mouat, Chief Scientist at Container Solutions. His blog post "10 Predictions for the Future of Computing or; the Inane Ramblings of our Chief Scientist" inspired this episode.
The post Day Two Cloud 120: Web Assembly, K8s Rivals, And Other Cloud Computing Trends appeared first on Packet Pushers.
Day Two Cloud 120: Web Assembly, K8s Rivals, And Other Cloud Computing Trends
On today's Day Two Cloud we talk trends and predictions in cloud computing, including emerging technologies such as Web assembly, rivals to Kubernetes, and the role of GitOps in infrastructure as code. Our guest is Adrian Mouat, Chief Scientist at Container Solutions. His blog post "10 Predictions for the Future of Computing or; the Inane Ramblings of our Chief Scientist" inspired this episode.Get to Know Your Tools: Nmap Ping Subnet Scan
Tony Fortunato of The Technology Firm uses Wireshark to observe how nmap discovers a subnet.Telegraf sFlow input plugin
The Telegraf agent is bundled with an SFlow Input Plugin for importing sFlow telemetry into the InfluxDB time series database. However, the plugin has major caveats that severely limit the value that can be derived from sFlow telemetry.
Currently only Flow Samples of Ethernet / IPv4 & IPv4 TCP & UDP headers are turned into metrics. Counters and other header samples are ignored.
Series Cardinality Warning
This plugin may produce a high number of series which, when not controlled for, will cause high load on your database.
InfluxDB 2.0 released describes how to use sFlow-RT to convert sFlow telemetry into useful InfluxDB metrics.
Using sFlow-RT overcomes the limitations of the Telegraf sFlow Input Plugin, making it possible to fully realize the value of sFlow monitoring:
- Counters are a major component of sFlow, efficiently streaming detailed network counters that would otherwise need to be polled via SNMP. Counter telemetry is ingested by sFlow-RT and used to compute an extensive set of Metrics that can be imported into InfluxDB.
- Flow Samples are fully decoded by sFlow-RT, yielding visibility that extends beyond the basic Ethernet / IPv4 / TCP / UDP header metrics supported by the Telegraf plugin to include ARP, ICMP, Continue reading