Going through the “Getting Started with P4” on AWS. Seems to work very well, but using a rather (free) weak compute node.
Originally written by Bruno Rijsman ; I was rather surprised there was no better guide, or rather a better tested Guide.
This exercise by itself is not very exciting, but it is obviously the first step to better understand P4 Programming and what it can do.
Will keep the blog updated as I progress
And there it is, the first update. Obviously the Free Tier Compute does not have enough storage to install all needed packages. Will have to use another resource for this exercise.
In 2005, I was a a Software team leader in Montilio, a company who tried (and managed) to accelerate File server access, by separating control plane from Data plane and making sure the data is not copied back and forth to CPU memory.
What I did not remember was that we were up against RDMA solutions, a technology I now promote. I still think Montilio had a real cool product, but it could not catch up to the fast pace of CPU and storage development.
Link to PR
And quote from the CEO:
Companies such as Alacritech and Neterion have attempted to alleviate this burden by using Remote Direct Memory Access and TCP Offload Engines, approaches Tsuk says are more designed to accelerate network traffic than make file processing more efficient.
As previously blogged, attended an OpenGov session focusing on AI .
It was very interesting to understand where different government organizations are at and how they interact with academia, what are their needs and future plans.
As this Blog is all about me, try to find me in two photos.
Oh well, I’ll help you with one.
Enjoyed participating in the Opengov breakfast insight session on AI.
Will wait for the official report and share it.
Took one photo while our photo was being taken, I like the composition and intensity of the photo so decided to share it
Finished the 3 first courses of the deep learning AI specialization .
Fourth and Fifth are not available yet, so not sure what to do? Cancel my subscription and wait for them to appear, or just wait patiently.
I really enjoyed the first two courses and exercises, including Tensor flow program, but the third one was less challenging.
Hopefully the fourth will have actual programming tasks for me to refresh my skills.
Now, I need to figure out a Machine Learning project that will exercise my newly acquired skills.
When I was in university, many years ago, AI was a big hype. Or so it seemed to me back then.
I obviously took AI classes, implemented Backgammon heuristics in LISP and lost the competition we had.
One of the projects (or was it a seminar) i did was to implement Neural networks to identify text from a small pixelated images.
Quite certain I still have a print-out of the code at my parents house, but I doubt if it make me proud readying it now.
And why do I write about this?
As I am taking the Deep Learning Specialization in Coursera, I implement Neural Networks in Python and in one hour of coding I achieve more than several weeks of coding, trial and error more than 20 years ago.
The following headline from Network Computing is accurate but misleading:
“Ethernet Switch Market: Cisco Declines, Arista Gains”
From just the headline one can assume Arista is getting close to Cisco Market share and about to gain the lead, but reading the article reveals more information.
Cisco is indeed declining, Market share dropped to 55.1% from 59%. This goes inline with global trend of having an option and Open Ethernet.
While Arista is gaining, 5.1% from 3.9% . An impressive gain, but no where near Cisco.
A more important detail and maybe a catchier headline would have been Huawei. 6.3% from 3.9%. Yes, Huawei surpassed Arista in Market share, yet the headline discusses Arista.
Not Fake news, but interesting interpretation.
Xcode installation is stuck in “Installing – 3 minutes” for more than 20 minutes, which gives me a reason to complain about Installation estimates.
We ALL know that there is no way to really estimate it , so why do it ?
It is helpful to know the percentage of installation and what else is needed to be done, but do I really want to state at a “Installing – 3 minutes” for more than 3 minutes? Moreover, it sometimes changes to 4 or 5 minutes, which is even more annoying.
As I was browsing Seoul map, I had hard time finding Gimpo Airport. Searching for it, I found it is a grey area with no details. Completely blurred out.
Was thinking it is some kind of security measure, but found out that the satellite imagery is very clear and detailed.
So, what is it ? Security ? mistake ?
Just for reference, you can see Japan Haneda airpot in Google map
Maybe apparent from my age, but I remember the times when programmers were working extra time to make sure their application “weight” is minimal.
Either memory consumption or actual binary size (for embedded systems with limited storage).
Was surprised (shocked) to see the new Google+ update is 111MB.
Yes, 111MB of application which shows content from the web.
Not a game, Google+.
I assume it is due to Retina display graphics, but I still think this is not the right way to do it.
Application can get the necessary files after loading, can’t it? Why everyone need to waste storage for unnecessary data ?