Is there a way to assign a specific (static) MAC address to a container, to allow dhcp assignment of the IP address from a physical router, while using macvlan?
yeah docker run --mac-address can do the trick

Do you have a tuo about how many hidden layer in a neurale network with tensorflow?
The optimal number of hidden layers is one of 'hyper parameters', a parameter which requires try-and-error for finding the best value. There's no theory to find the best hyper parameters for each neural network, so every ML researchers are doing try-and-errors with various different combinations of hyper parameters. To mitigate the burden of hyper parameter tuning, Google provide HyperTune feature in our ML Engine product. The tool support you to find the best hyper parameters, such as number of hidden layers, with smarter way (Gaussian process) of trying out. (https://cloud.google.com/ml-engine/docs/how-tos/using-hyperparameter-tuning)

you think that machine learning will change how people write applications or those machines will only limited to particular cases like image recognition or data prediction?
Machine Learning is not a silver bullet, but would have significant impact on various applications including programming. I'd imagine that future IDEs would have a code assist feature powered by machine learning, suggesting possible code snippets or refactoring ideas based on past practice or patterns, for example.

Tensorflow is Aldo available on premises or just in Google platform?
TensorFlow is an open source tool that could be run on any PC or cloud platform. At Google Cloud, we provide a fully managed version of TensorFlow as ML Engine. (https://cloud.google.com/ml-engine/)

Which is the best starting point for learning about TF specifically on written natural language recognition?
There are many approaches to apply neural network to Natural Language Processing, but one popular algorithm is word2vec. It may be interesting to learn the technology as starting point. (https://www.tensorflow.org/tutorials/word2vec)

In Image tagging using cnn how the convolution matrix where defined?
We are continuously improving the Amazon Rekognition training, for more information you may have a look at this presentation: https://www.youtube.com/watch?v=fk-TxySUAzw

Does lex work with Italian language?Is there something to understand written natural language?
Currently Amazon Lex is in preview and supports US English only. We are working on adding more languages (based on customers demand) as soon as possible.

When italian language will be supported by Amazon Lex?
Currently Amazon Lex is in preview and supports US English only. We are working on adding more languages (based on customers demand) as soon as possible.

Probably machine learning will be part of all things in few years. Do you think that people has the right perception of this at today?
I think we need a greater awarness about the possibilities that machine learning and artificial intelligence are opening. Developers and technical communities should be in the front line for that.

How can we choose the right number of layers and the number of perceptros per layer in order to solve a specific problem through a neural network ? How do they affect the output of the network ?
Finding the “hyperparameters” that configure a model, such as the neural network architecture to use, is an open problem. You can iterate and test multiple values, or use some heuristics. For some hints, you may look at this interesting online training: http://course.fast.ai

Cloud providers as AWS are planning other services based on machine learning? As developers we will have to deploy mostly with machine learning based services?
AWS wants to Brings powerful artificial intelligence to all developers. We offers a family of AI services that provide cloud-native machine learning and deep learning technologies to address your different use cases and needs. Over time, it will be easier and easier to use high-level AI services, or platforms/engines that you can train with your own data.

Does the function have a cold or an hot start?
The first time a Lambda function is invoked, a container is provided to execute the function. You may keep a container warm scheduling an invocation - for example - every 5 minutes. You can use CloudWatch Events for that.

Is it possible to use web sockets with Amazon Lambda?
Not directly, but you can connect to AWS IoT using WebSockets and then messages send over WebSockets can trigger (using an AWS IoT Rule) a Lambda functions. For more information, you may look here: https://aws.amazon.com/iot-platform/how-it-works/

How to keep the focus on the whole application while having maybe hundreds of functions?
In my book I stress on how to keep an event-driven design, so that the complexity of the interactions is kept “local” and you don’t have to think of a complete workflow of all functions (choreography is better than orchestration at scale). AWS has also release multiple features to make this easier, for example: - AWS Serverless Application Model (AWS SAM) provides a simplified way of defining the Amazon API Gateway APIs, AWS Lambda functions, and Amazon DynamoDB tables needed by your serverless application. AWS SAM is made available under the Apache 2.0 license, enabling others in the ecosystem to adopt and incorporate it into build, deployment, monitoring and management tools with a commercial-friendly license. https://aws.amazon.com/about-aws/whats-new/2016/11/introducing-the-aws-serverless-application-model/ - AWS Step Functions makes it easy to coordinate the components of distributed applications and microservices using visual workflows https://aws.amazon.com/step-functions/

Why a Lambda function need to recall themselves? Why doesn't use queue system like SQS or similar?
You can invoke a function directly of via a queuing/notifications service. For a direct invocation, I don’t see value in going through another service.

When AWS Lambda start support Python 3.*?
We are working on updating our runtimes, and supporting new ones.

Which are the best tools to test and code locally serverless applications?
In my book I show how easy it is to wrap a function to execute it locally. Most frameworks have some kind of automatic support for local executions, for example you can look at Chalice: https://github.com/awslabs/chalice My personal suggestion is don’t develop locally unless you really need it: you can now use the same environment for dev, test and production, with the same performance, scalability and latency, without any noticeable costs (see the Lambda free tier).

Which protocol is used to ask the device for a reboot (from server to client)?
Direct methods used, for example, to start a reboot, are HTTP-only from the cloud side and MQTT-only from the device side.

Could you please explain the main differences (a part from pricing) between event hub and service bus? they started as part of the same service but it looks like now have been splitted into 2 separate ones
The main difference is related to the scaling capability: with Service Bus Queues and Topics each message has a TTL and receivers are responsible to remove messages when they do not need them anymore. Event Hubs has different partitions to be able to receive a huge amount of messages, and several consumer groups can be defined, to allow multiple receivers to read concurrently the same messages at different rates. Each receiver has its own pointer and messages are not deleted by receivers, but there is a common retention policy for all. Azure IoT Hub is the evolution of Event Hub, with support for device management, bidirectional communication between cloud and devices, MQTT protocol.

Is it possible to analyze the ingested data with some big data tool? any tool from Microsoft?
There are several options to analyze data ingested using IoT Hub, here just few of them: Azure Stream Analytics for real time stream manipulation, Azure HDInsight is a cloud distribution on Microsoft Azure of the Apache Hadoop technology stack for big data analysis, Azure Data Lake Analytics is an on-demand analytics job service to transform big data and extract insights, Azure Machine Learning to use ingested data to create models able to predict specific results.

If the manager is compromised, can the attacker read all the secrets? How can we remediate?
The managers work as the secret store, so they'll have unencrypted secrets in memory. If someone compromises the managers, you'll have to remove the compromised manager, rotate the Certificate Authority (one command), and use docker secrets to rotate all of the compromised secrets.

can use docker engine in a selinux env?
Docker engine supports SELinux, and ships with a configuration by default already.

If. Security swarm Can work in multiple datacenter in public network?
Yes, but I would recommend either using different swarms for each datacenter, or at least, have all the managers deployed in the same datacenter.

We're moving the security problem from infrastructure to a single central element : the docker-engine itself. How we can be guaranteed that a bug/exploit in docker-engine doesn't compromise the whole security of our system?
This is the point of following a least-privilege mode: compromise of a node only compromises what the node has access to, no more, no less. In terms of guarantees, we use security best-practices around code development/review and do external audits of our code to ensure everything is done correctly.

In order to deploy 6+ times/day how much do you spend in QA or UAT? Did you fully automated testing?
Testing is key to success when handling complexity. Every piece of code we deploy to production includes unit, acceptance, integration tests. When necessary, we have performance and security tests too. Everything is automated so that every commit to github triggers the entire tests to run and ultimately to deploy without human intervention.

How do you keep UX consistent across dofferent components? And what about stylesheets?
Keeping consistency in very hard. Most of what we do is to evangelise good use of design patterns for behaviour, and styleguides for styling. Nevertheless, we trust developers are going to do the best for the end user if we make easy to do the right thing, for instance by modularising our code and by keeping our good balance between DRY and OPD (optimise for destroyablity, aka KISS in the microservices world).

How much it cost a simulation like the one you did?
Less than 0.50 euros

In your opinion a well organized scalable layer (with autoscaling) is better or not that a serverless architecture (both hosted on cloud) to manage intense traffic?
Serverless architecture vs PaaS (or IaaS with autoscaling) is a complex topic that opens the door to several discussions. I strongly suggested you to have a read to this amazing Martin Fowler' article (https://martinfowler.com/articles/serverless.html). From a general point of view I could say that it depends by the use case: if you need to implement a quick and simple service (without using 'esothic' third party libraries) I think that serverless approach could be a good fit; if you need instead to have more granular control over the servers serving your requests IaaS/ PaaS will be the perfect choice.

Having multiple databases causes data replication, which causes data to occupy much more space, and have a much higher probability of being misaligned, my data is misaligned compared to the other systems of the service chain (assume the web portal, the bank gateway, the billing platform) and I have just one db!
I agree that data replication is an issue, even more in a microservices architecture where the relative independence of each service can lead to a sort of unmanageable anarchy. That's the reason why one should take extra care in designing its data layer so that it reflects the separation of concerns that is proper of the microservices approach: each service should manage all and only the data it is responsible for.

Is a good idea in your opinion mix microservices from different providers?
Microservices are indeed a very good choice when designing a portable solution. Due to their independence wrt the underlying architecture, you can choose to deploy the same set of services to different platforms just to implement a HA solution, or you can 'mix and match' different providers using the BaaS tools and technologies that best suit your needs.

Basically Google made a product in response tu aws lambda. What are the main differences and which One is cheaper?
While the comparison is not completely fair, as AWS Lambda is for sure a more mature product (and includes support for additional languages like Python or Java), I would say that a good feature of Google Cloud Functions is the just-in-time execution of npm install that allows you to deploy your function without generating first a ZIP package, and allow a tighter integration with a GIT versioning system. In most use cases GCF are cheaper than AWS Lambda (and they have a 2-million invocations free tier, while AWS has 1-million free invocations).

which of these dev tools are already available and usable with Firebase functions SDK?
I assume here you are referring to the cloud tools like Cloud Logging and Error Reporting: Firebase Functions are nothing different from 'native' Cloud Functions, so you basically have access to the same entire set of dev tools.

Is the support of other programming languages like golang or dart coming to your FaaS in the near future?
I'm not a Googler so I don't have 'first-hand news'; however, GCF PMs confirmed that other languages are in the roadmap.

How about distributed transactions in a microservice architecture?
This is a very interesting topic, which is way broader than the scope of this short reply :) and even crosses the boundaries of specific technologies like Google Cloud Functions or AWS Lambda. Maybe you are already aware of that, but I'd like to point you to this Medium post (https://medium.com/@KevinHoffman/distributed-transactions-in-a-cloud-native-microservice-world-7528f8baa8da) which is a good entry point to the main keywords that pop up when talking about transactions AND microservices (namely: Single Responsibility Principle and eventual consistency). This is a 2015 post, just to give an idea of how long this topic is being discussed.

Do you think that Cloud computing is moving to frontend devlopment?
I think the Cloud can fill in a lot of server side development for developers. Particularly features in applications that aren't the core business and that you don't want to spend time on. If you have the APIs and the JS SDK available, and increasingly you do, then you only need to concentrate on building the interface the way you want.

All Twilio Services are available in Italy?
You can purchase Twilio phone numbers in Italy to make voice calls (https://www.twilio.com/docs/api/rest/making-calls). You can send SMS messages from international numbers or using an alphanumeric sender ID (https://support.twilio.com/hc/en-us/articles/223181348-Getting-started-with-Alphanumeric-Sender-ID). Any of our services that don't require telephony are available too, so you can also check out and use the betas of Programmable Video (https://www.twilio.com/docs/api/video) and Programmable Chat (https://www.twilio.com/docs/api/chat). Also keep an eye on Notify (https://www.twilio.com/docs/api/notifications) and Sync (https://www.twilio.com/sync/api), which will be really useful building blocks for applications.

API drive applications are rising. Do you think that PHP development is moving to only API development?
I don't think PHP will be only used for API development. For sure, I see a growth of API development in PHP, especially now thanks to the performance speed of PHP

How do you see PHP in the next future, also considering the huge growth of cloud computing and hype (sic) of serverless approach (i.e.h AWS Labmda functions handle just JavaScript, python and Java scripts)?
I see PHP in a very good fit for cloud and serverless architectures. Using a middleware approach, we can build brilliant microservices in PHP 7 that are fast and easy to manage (a PHP application is very easy to deploy in production and the PHP engine runs everywhere).

The dockerized is also available on platform like aruba private cloud?
The driver doesn't support the servers on the Private Cloud, but you can use the Attach command to manage existing servers with docker. The only requirement is that ssh is installed on the server. In this scenario, please note that all the commands that use public API of Aruba Cloud won't work. We are considering to improve the driver in order to support Private Cloud.

It is possible also to use the new docker service command?
Yes, the Service command is supported because of it is a standard Docker command. It is required that local docker and remote docker are both up-to-date with a release supporting the command. To update the remote server is enough to use command update of docker-machine.

Are you planning to introduce new sdks for other languages like PHP?
Both the Cloudant and Watson Alchemy APIs are simple HTTP. There are languages and bindings for lots of languages but you can simply make HTTP requests to start using these APIs without any libraries

Watson will support also video analyze?
At the moment, Watson's image recognition API is on static images. I have seen developers extract frames from a video every second or so and then analyze each frame to produce a searchable database of what a video contains.

Is a goog choise mx watson with other serverless products for different providers in the same application?
Cloud services allow you to pick and choose the services that work best for you from whichever supplier to need to use. In some cases, it makes sense to have your services near each other (for low latency, for example), but developers have the flexibility to compose systems from a number of suppliers because of the easy-of-use of cloud tools and the low or zero cost to try things out.