Translated from 日本語 • View original Yes. Now, here is the second part of the flight (to San Francisco). The material is too old, but don't worry about it.
Since our company was participating as a partner, we participated from the Partner Summit held the day before GCNext. To summarize the content, Google Cloud is a good product, so please sell it. I have no choice but to sell this.
After finishing the partner summit while being half absorbed in the atmosphere of the venue, it was time for a party in the adjacent park. The venue was large, but there were more people than that, so we naturally shared a seat, but when I talked to the handsome guy I met with the help of beer, he said that he had come from Italy for 12 hours. Suge!
When I honestly said it was amazing, he replied, "No, you too." I see. We also came from the other side of the world for 10 hours. It's crazy.
I was strangely moved by the fact that there are so many Google Cloud fans from all over the world, including us, who come all the way to San Francisco, and people who want to do business. We're all together!
It's a small world!
Now, let's take a look at the long-awaited announcement from 4th to 1st place. I feel that the ratio of the splits is a little strange, but I will proceed without worrying about it.
#4: CloudSQL for MySQL Live Migration Capabilities
This is not a keynote, but a session material I attended. I didn't know about PostgreSQL recently, but it seems that CloudSQL for MySQL supports live migration from MySQL running on-premises and other clouds. No, I really didn't know.
When trying to migrate an on-premise system to the cloud, it is human nature to want to reduce downtime, preferably zero. If the data store stores store a large amount of data can be migrated without downtime, it can be said that the way to migrate the entire system to the cloud without downtime has been opened. In the session, we talked not only about product-specific but also about know-how to reduce downtime in general.
According to him, data migration consists of exporting and importing data, but what is noteworthy is the import time. There are three strategies to shorten import time.
Make the import destination instance type as large as possible After the import is completed, you can make it the appropriate size.
Add indexes later except for absolutely necessary indexes Reduce index build time.
Stop logging that is not absolutely necessary during import and start after the import is complete.
I see. In particular, keeping the instance type large is a solution unique to the cloud.
In addition, at the last question time of the session, will there be a live migration to PostgreSQL? There was a person who asked. Thinking is the same all over the world. The answer is
YES.
It was. I want you to come as soon as possible.
3rd place: BQML impact
BigQuery Machine Learning, or BQML for short. What! It is a service that allows ML to be done with BigQuery.
This Youtube is easy to understand. The short time of 3 minutes and 43 seconds is also a high point.
As with AutoML, ML still has a high threshold. I can feel Google's willingness to lower the threshold to a height that ordinary engineers can straddle. Many companies are putting data into BigQuery for analysis, and you can create training models on top of that very BigQuery without having to move that data elsewhere. Analysis results can also be obtained. Moreover, it is quite a shock to know that it is a query with SQL-like grammar.
If ML can be done using BigQuery, a data store that has a reputation for ease of use, I feel that the scope of use will expand at once. This is a must-check!
2nd place: KNative
KNative's K is K! for Kubernetes, and it is a service to realize things represented by Serverless and AWS Lambda, which are becoming a bit of a buzzword, on a container-based basis.
If I were to explain what makes me happy before I have used it, it seems to make container operation in kubernetes, which I think is still too difficult for mankind, easier and easier, and that is exactly what makes it as good as Lambda. I myself am still learning Kubernetes and have no track record of putting it into production, but I still feel the complexity at this point. Around Ingress.
The strongest Serverless platform? I also saw the appearance of Knatative . Is it because of my mind that I don't feel like it's getting so easy? At the moment, I don't have the same enthusiasm, but to be honest, I want it to be easier. I hope that containers will become easier to handle.
As a product that is also talked about at the same time, - Istio - Apigee
There was. The understanding is that Istio is an integrated approach to microservices running on different platforms, while Apige is an integrated approach to APIs running on different platforms.
If you somehow diagram the stack,
[Istio]------------+-------------+
| | |
[Knative] | |
| | |
[kubenetes] [apigee] [other micro services]
|
[Docker]
|
[GKE vm cluster]
Is it like this? The stack is quite complex, so it is an image suitable for enterprises. I will continue to watch.
1st place: Ni-chan, a black man who did a live demo
Keynote 3 days, from around 25 minutes and 45 seconds.
Please watch it. Ni-chan, Kelsey Hightower. If you've ever done a live demonstration of this complex procedure, you can calmly complete a demo where you can see that it will fail due to network errors or resource problems along the way! I'm numb and longing for it!
So, it's amazing.
The jokes are also quite black and funny.
If you haven't heard of StackOverFlow. You have no clue what's going on. This is high quality code that you can find on this thing,man. Trust me. Or, there is no engineer who doesn't use StackOverFlow www It's common sense that you shouldn't trust the code there, www The person who asked the code that came up after searching for it is the person who is doing the demo or is already completely prepared wwww
StackOverFlow post in question
If you're an engineer like this, you can't help but laugh. So, of course, the venue was bursting into laughter. I was able to gain a valuable realization that StackOverFlow material is common all over the world.
Next
So one thing is, "Good programmers copy". Great programers paste.
Copy and paste programming wwww
Once again, the audience burst into laughter.
However, I thought it was just a story, but I created a table in Postgres and poured the data in a flowing manner. Next, launch the API container on the Kubernetes cluster.
Some of the components of the Kubernetes cluster are said to be hanging on to other servers via Envoy. Whether it's GKE-on-prem or the power of Istio, it may be natural for a Google super engineer, but it's numb to be shown something with a complex structure that seems to be so simple. I wonder if the explanation of DaemonSet is also a point in the case of via envoy. By the way, DaemonSet has a limit of only one pod being launched per node.
Perhaps because the machine I was presenting on was a Pixelbook, I left all the container builds to Cloud Build. He said that he did not install Docker locally. No, in the first place, there is no Docker for Chrome OS www
In an instant, CronJob, which retrieves weather information via API and stores it in the DB on a regular basis, was followed by the front-end API server. Connect with OpenCensus and monitor metrics with StackDriver.
The voice UI was cool, so I quickly connected it to the API server running on GKE with Dialogflow and finally started testing the operation.
**_Google Assistant:
Looks like you have pretty large crowd here. I hope the demo gods are on your side.
Kelsey:
_**
I hope so too!
Even if you know it's already prepared, you can't help but be amazed at the result.
When I asked about the current temperature in Los Angeles, it gave a perfect answer and proved that the pod was working properly by spitting out logs.
Finally, the Assistant is
Okay, I gotta admit. That was pretty dope.
Clap, clap, clap. It was a wonderful demonstration.
Someday, I would like to do a live demonstration like this. Nine times out of ten, I will fail...
So dogma and prejudice are severe, but it was the top 10 topics on GCNext.
In addition, GCPUG Hiroshima (sorry if I'm wrong!) I met people who were supposed to be unlucky, and I was keenly aware of my lack of English skills every time I tried to chat with Luck on Uber, where I shouldn't have to talk about anything. I would like to introduce this area if I have the opportunity.
**See you next year at GCNext. **
[
](https://www.colorkrew.com/gcp/?utm_source=gcnext18_2&utm_medium=referral&utm_campaign=owned)