 Tim Seller in my team has this great saying that when he doesn't call me a manager he calls me a host Which is very true because we're just all engineers in Google and we're all engineers here in this room, right? So I think that's more important for me than anything else Lots of you have actually already seen me before. I know lots of familiar faces here in the room So that's why I've presented a slightly different talk today And you see that already in the in the title It says my favorite announcements from from cloud next There's two reasons why why I put this my in here The first one is I know all of you know GCP and no Google cloud probably many have watched the live stream or anything So I just wanted to give you a little bit more personal view what I found like interesting rather than giving you the official rundown by by importance But the second thing was also that I want to make this session very very interactive and with my I wanted to also invite all of you to a dialogue We have officially 45 minutes But they have quite a few of videos at the end from demos that I that I personally found pretty cool But I can skip all of them. So let's please yeah raise your hand ask questions anytime in the middle Let's just make this a session between all of us. Okay, this doesn't have to be a lecture So please just raise your hand. Yeah, let's let's talk. We are 45 minutes and There is there is no great conclusion at the end or something else that I'm going to present to you So we can at any time slow down speak up again. I'm very I'm very happy to do anything. Okay, cool Awesome any questions yet? Maybe regarding the fire instructions Okay, no, okay, great. So, yeah, let's go right in so I was actually a club next in San Francisco If you have never been I can only suggest to go It's completely unbelievable because we are basically taking over a whole square mile of San Francisco in the middle of the city It's pretty amazing What stood out for me was last year. It was a lot about AI and ML and it was a lot of product focus last year This year changed a little bit and when I explain it I usually say for me it changed from verticals like products to horizontals suddenly everything is integrated And so we see that for instance that there was a lot of focus on openness and Google being the open cloud and the open source Cloud there was a lot of focus on security which also again cuts through all of the products a lot of focus on hybrid Working with other clouds, but also working on premise and that's what the least Integration between our products so rather for instance later is giving a pretty cool talk how big query and data flow are now integrated So that was really a very interesting change somehow for me and I think also something that's very important for enterprises and generally for all kinds of companies because that's what we All deal with every day. I used to be a software engineer and a data integration architect myself And it's actually the connections that are the hard parts That's what people always say in architecture diagrams. Don't look at the boxes. Look at the arrows And so and that was something that I really really liked about club next So in that spirit, I'm going to go through a few of the major announcement and what I found interesting about that again At any point in time raise your hand or if you have comments that would be even better if you say I've used this I don't like it because of blah That's all would be awesome Okay, cool. So start with infrastructure on hybrid. Well, we've announced two new regions in Korea and in Utah That's in the United States of America Bring in the total to 23. I'm gonna have the next slide with a updated map in case you're interested Then we have Anthos. I'm gonna have Anthos later I'm gonna go specifically over that but the one liner is it's our cloud services platform And now rebranded as a complete suite for for hybrid service solutions I think it's it's pretty exciting, but it's also really really complex. So that's what I'm going to talk about that a little bit later Then we have new VMs in particular compute and memory optimized VMs I'm not sure who has seen the keynote But there was a really cool demo of the live migration of SAP HANA. So if you don't know what SAP HANA is SAP as an enterprise System for like everything enterprise basically and HANA is an in-memory database and that runs now on those memory optimized VMs So they have especially large and fast memory attached And we were doing a live migration basically shutting down certain instances and you could see that subcontinued Progressing transactions as exactly the same speed. It didn't lose a single transaction while the HANA migration was happening So that was pretty cool What's interesting is you can now also use those VMs on GKE on Kubernetes And we also changed the UI so I know I guess none of you ever interacts with the UI with the Google Cloud console I do neither I like scripting but nevertheless maybe have a look because the few assistants actually changed So now if you start a GKE cluster, you can actually choose what you need and it's nicely written in like terms that everyone understands Like I want to optimize for machine learning I want to optimize for SAP and then you can choose your clusters depending on that and that's possible because now we have those optimized VMs And last but not least That's something very interesting for enterprises BYOL that means bring your own license with sole tenant nodes So you know there's a few enterprise products out there which have CPU based licenses And that in the past used to be a problem if you're running on VMs because what is the physical CPU? Well now we have sole talent nodes which means we can ensure only this VM is running on the actual hardware for licensing reasons And you can bring your own license so you don't have to purchase a license And that's very important for customers who have for instance large contracts with I don't know Microsoft Oracle and so on Where you don't just pay for like one license, you don't want to pay that via Google You have a massive enterprise contract that somehow covers all kinds of licenses in a model that we can't map otherwise So now you can bring your own license and run that on sole tenant nodes which is super cool And we announced that as well Any kind of questions for those announcements? Am I going too fast? Am I talking too fast as usual? Ok faster, ok Look lots of people say that I talk too fast so I try to be somewhere in the middle Ok cool I promised you the map this is the last map now the post next map of all of the regions that we have in the world And here are the new ones so I just mentioned sole and Utah is now officially on the map And you also see Jakarta here Jakarta is going to be launched I think later this year beginning next year What's interesting about Jakarta and that's why I have put the next slide on here of the upcoming network connections Not sure if you knew that but there is Indigo going online this year which is the largest undersea cable in the whole region And that's going to go via Jakarta and of course via Singapore So I think if you have those kind of workloads super interesting Because yeah I guess you can expect much much reduced latency in the future due to Indigo going from Sydney to Perth to Jakarta to Singapore and then even up And that's a collaboration between Google and between I think Telstra in Australia between SingTel and between what is it called in Jakarta Indosat or something like this Indotel So that's super interesting and very exciting for me I like networking stuff and it's pretty cool to see how we're now integrating Asia as well I think in the past we didn't see much green in this area and now it's really really dense so that's really good to see Yes, sorry? What does it mean for SGC, Japan, Hong Kong and Singapore? Oh those are the names of the undersea cables so you can literally go to Wikipedia and there will be an article about that's just a name This is called Indigo, SGC is the name of the existing undersea cable that already connects Japan, Hong Kong, Singapore since 2013 Which is one of the reasons why for instance if we have multi-region replication for high availability that's why we always suggest Singapore, Hong Kong is a good connection Because we have a dedicated line there Total number of regions 23 right? Did you just count them? The question is on the architect exams what you should answer Well I think it's changing so much, the good news is it's changing so much that the documentation is inconsistent isn't that great? Sorry So you don't know what's the right answer? I can calculate now again I just literally this morning copied it out of the documentation I hope 23 is still correct Most likely the number will be higher if in doubt always choose the higher number Okay cool I'm going onwards to serverless and ops So first of all GKE On-Prem who has heard about GKE On-Prem already that was actually pre-announced last year Cool so what's good about this is this is now officially in beta so you can actually request people like professional services Yours truly to work with you to actually get GKE On-Prem installation and we have actually installations running in JPEG already Just for the ones of you who don't know what that is That means we run Kubernetes for you on-premise so on your infrastructure that means on your VMware We install a Google managed version of Kubernetes that is exactly the same as GKE and is connected to GKE So you can manage your workloads in the cloud console but have your on-premise GKE cluster running We sometimes call that running on the edge right so you have for instance certain type of data that you have to keep in your data center Or you have an application that for some strange requirements has to be in your data center For instance I worked on a project that was a telephony application that needed some special hardware Fair enough that can't move to the public cloud but you can now use GKE On-Prem to only have the edge components running in your data center And connect it to GKE on the Google cloud and have it all in one single pane of glass so that's pretty cool Yes, yes Question is first of all is there a hardware requirement or you guys run everything on VMware? It's all VMware at the moment yeah, so there's no hardware So as long as you are running on certified hardware then it is supported Exactly, so we actually have a partnership with VMware there who then again also check the underlying hardware The question is the GKE means that there is no stretched cluster and all the worker nodes are on the frame rather than in half here and half there Yes, so just to repeat that question, sorry the first question that you asked was can I run it on hardware or is it always VMware? And my answer was it's VMware and the second question was does that mean that I can split one cluster into on-prem and into cloud or is it two different clusters? At the moment it's two different clusters having said that because we soon bring Anthos and the service mesh this idea of a cluster becomes a little bit more fluid Right because clusters can scale between different regions and clusters with Anthos Soon as you if you run a service mesh then you don't really care about the cluster anymore because we manage the clusters for you It's all about workloads and services but yes at the moment it's two clusters that are connected For service mesh you mean Istio, right? Yes, exactly So Anthos is more than that, that's what I'm saying, like Istio is kind of a complementary to Anthos But not all the features of the cloud are supported on-prem, right? So you cannot attach GPUs or do you support support? So the question was not all of the features that are on the cloud are supported on-prem That's correct because in the end we don't know your hardware so there is a choice between For instance, we don't have spot instances or preemptible instances, those kind of concepts we don't have Because we can't control your hardware on-prem So normally you would have a fixed cluster that you dedicate for your GKE on-prem installation That's actually a good point, I would be interested about the use case Let's maybe chat later over pizza Cool, okay, I'll continue Cloud Run, what's Cloud Run? Cloud Run used to be called in the past, well it had different names But this is basically you would know it as Knative So that is the idea that you run Cloud Functions but run Cloud Functions on Kubernetes And you can decide yourself where you run them on This whole hybrid really coming through both with GKE on-prem and with Cloud Run Because our position is that just because you want to use functions you shouldn't use a different framework for that So if you use like another serverless function framework you basically have a completely different development stack And the idea of Cloud Run is that actually you should use the exact same code It shouldn't matter whether you want to deploy it as a function and pay it for every call Or you want to deploy it on your existing GKE cluster and pay for the underlying infrastructure for the runtime This is a decision that someone else should make, not you as a developer And that's where Cloud Run comes in So you can now choose completely freely using Knative on all open source components as usual with Google And how you want to run your workload So it's real DevOps, right, you as a developer you don't work around Ops, you work together with Ops to find out what you actually want to run And how do you want to run it, that's really nice Nevertheless we still had a few changes to App Engine actually We have new runtimes and we have a private GCP access there And what I want to particularly call out is Cloud Code Why do I want to call that out? This is a plugin for IntelliJ and VS Code So we admit that VS Code is pretty cool, I use VS Code myself So why not leverage it? And yeah, so we wrote actually a plugin for VS Code That you can now use all of those functionalities For instance the libraries for Cloud Run directly from VS Code So you can write a function in VS Code And directly deploy it as a function into GCP out of VS Code Just completely seamless But at the same time you can still from VS Code decide Oh I just want to run it on my existing GKE cluster Or I want to run it on GKE on-prem And all of that is possible with just like one click And again, all open source and all transparency So that you know what's actually happening behind the scenes Okay, networking Traffic Director, I'm going to talk about that later Traffic Director is basically a service mesh That doesn't require services That's an interesting way of explaining it That means we can use a proxy Like for instance like an NVR proxy But we provide a control plane That configures all of your proxies In a way that you can control the overall network traffic Even if you're not using containers And even if you're not service based So we also use Traffic Director internally If you use services and if you use containers But with Traffic Director we also offer you the same If you're not doing that Which is again like a very open approach Then we have private Google access That is now launched That's not generally available That's something especially interesting for enterprises You can basically disable that from your GCP You can access anything else So you get private API endpoints You can literally even block the DNS access So you cannot go outside of the cloud And also you cannot connect to your services From on-prem somewhere else This normally is not a problem Because you have IAM permissions But this reduces the insider problem So what do you do with someone who gets somehow rogue permissions That someone just escalated And now that person has permissions they shouldn't have But they still have access to your systems And with that you can reduce this So that person could still only access your front end And then nothing else on GCP Because you have defined parameters We also now offer 100 GB interconnect That's a lot That's per second by the way Just in case you were wondering And high availability VPN We always used to have a solution for high availability VPN But you had to configure that manually It was available but you had to know what you're doing And now you can just directly set up high availability VPN Okay, I'm continuing to security I just mentioned the private API access If you want to bring that even one level down We have VPC service controls now I actually worked with a large banking client Almost last year to introduce that That really brings it down to the network level So you can exactly say for instance From BigQuery if you're exporting data It can only go to this bucket And only this group of users is allowed to export it to this bucket So very very very fine granular permissions Again something that normally You should solve with groups and role permissions But if you have some kind of an insider risk Or you're not really sure who uses your service account In which way and your applications You can put this layer on top to make sure That there's no privilege escalation possible either Something I want to talk about later Is the Cloud Security Command Center And the Data Loss Prevention UI So now in the UI you have a lot more control Over the security across all of your applications And across all of your projects There's something that lots of enterprises have asked for Because they have central security teams And they don't care about the applications They don't care about your DevOps teams They are a separate security team And they want to check the security For everything across all of your applications And you can do that now with Cloud Security Command Center We also have scanners for the container registry And for VMs as well So all kinds of vulnerability scanners included And also authorization So after you scan something And after you have basically provided evidence That this is a golden image You can actually tag that And then only allow this to be deployed And with shielded VMs We go even one step further Once you deploy that We have even a runtime firmware That uses hardware modules To ensure that only this is run And it's called a shielded VM One last thing That's very often requested And finally here is Microsoft's We manage service That means now you can actually If you don't want to use Cloud Identity You can still use your X directory And can actually sync that to identities And that's as a managed service So there's nothing you have to do yourself Okay, I'm going a little bit fast here But again I'm always happy for questions So I'm quickly pausing So just now we were talking about the cloud one Will we eat one on-prem? So the question was Does Cloud Run work on-prem? So Cloud Run itself is a cloud product But everything under it runs on-prem So you can either run it in GKE on-prem Or you can just use Knative yourself But of course the whole idea behind Serverless is that you don't have to manage it So if you want to run everything on-prem You most likely You don't get the full benefit out of Cloud Run That's how I would put it Right, because you want Not even to know where it's running Is it part of the end software way? The question was whether Is it end software? No, it's in parallel to that It can work together with Anthos So Anthos and Cloud Run are very tightly integrated But you don't have to use Anthos If you want to use Cloud Run So Cloud Run is really more for the person Who develops functions Like Lambda functions But still wants to have to control Over the infrastructure behind the functions Which not all other serverless offerings Allow you to do Okay, cool, storage We have a new storage class announced It's called archive storage Sometimes called ice cold storage And this is basically to replace To replace your tape backups So this is a very, very Cheap storage class But that still allows you fairly short latency Requirements In the announcement blog post And this is a quote We've wrote it's not as glacially slow As other solutions Make up what that means yourself But yeah, it's basically So it's still on hard drives But it provides you all of the features That you expect from a type backup solution Especially in terms of cost So that's a very interesting new offering We also have file store That's our NFS offering Is now in NGA And also provides now High availability and multi-regional Replication if you need NFS storage And then we have a few interesting Tiny changes that I still find Somehow intriguing So we have a bucket policy That means you can now define permissions On a bucket level Not only for IRM groups themselves So you can define different permissions Even on objects inside buckets Similar to ACLs And if you have your own Signatures coming in We now also support you to sign Buckets and objects as well And I mentioned here Big query Going a little bit into the data direction If you never heard of BigQuery data transfer service This is an offering from us Where we copy your data over And one of the big announcements of NEXT That actually kind of went below the radar Was that now we have over 100 integrators For BigQuery data transfer service So where we have a connector To your on-premise solution For instance, teradata That migrates the data into BigQuery Via Google Cloud Storage Typically exported in Avro Or Parquet or one of those formats With functionality under storage So it goes via Cloud Storage into BigQuery Okay, let me quickly have a look at the clock We're still doing fine Data analytics, finally my area No one asked me too hard infrastructure questions And not actually infrastructure person So thanks for not doing that Now there's the area where I consider myself At least half an expert in So please ask me questions here Connected sheets, has anyone heard about that? Great, so we're going to have a small video On that later because it's actually pretty cool Other question, has anyone of you Used Google Sheets before? Come on, send that Google Sheets G Suite, look up from your phones Exactly, so there is a limitation Similar to Excel as well That you can have a maximum number of rows In that, it's a large number of rows But nevertheless there is a limitation I reached the end, right? I reached the end of Excel Exactly, you reached the end of Excel I'm not sure if there's always a good news Anyways, so the connected sheets They basically erase that limit So dynamically in the background Once you reach a high number of rows We switch over to BigQuery You don't even notice that So there's nothing you have to configure We just basically plug in temporary BigQuery And suddenly your Excel sheet Gets like a turbocharger So it's literally like Turbochargers used to work in cars I'm not sure if anyone of you know how they work But basically the first RPMs It doesn't do anything And then there's a point when they click It kicks in So that's exactly the same thing You use sheets as long as it's powerful In the browser And then at some point under the hood Bam, BigQuery comes in And then you're getting really fast So that's really, really cool And I have a short video on that later Why don't you put the BigQuery at the beginning? Sometimes people don't want that The question was Why don't we pick BigQuery at the beginning? Fair point I would also do that But lots of people come from an Excel background Or come from a sheet's background So they just like sheets And to be fair There's one valid point That I always bring is That the sheets is really nice for collaboration If you have multiple people working together And actually that's the video I'm showing later How AirAsia completely changed Their internal meeting culture Because now the numbers are just always there So Nick Kunsch from AirAsia is talking about How they used to always have two meetings One meeting to talk about the numbers And then in the middle The data scientists had to get the new numbers And then a second meeting to talk about the new numbers And now they just do this within the meetings So they just literally say 50% of all of the meetings Which is nice I was like okay This is a big number of meetings Can you help them to fix their booking engine? The question was Can I help them to fix their booking engine? I don't know anything about the specifics I particularly like AirAsia So yeah, but look I have a lot of problems with booking Singapore Airlines And look under these days So maybe all airlines Have their own problems It's also a complicated business So we should give them kudos for that Data fusion That's our point and click Like UIs, ETL generator So you can really Just similar to what for instance Informatica Or IBM data stage offer you So it's a UI Where you can define your own ETL pipelines At the moment we support Spark But soon we're going to support Dataflow and Beam as well So it creates those pipelines for you Based on a UI experience That's the kind of thing, it's awesome Yes, sorry How about collaborating in this Is it under data fusion For the booking Salesforce collaboration The question was Google and Salesforce Collaboration under data fusion That's a good question I'm not sure if data fusion has a Salesforce connector I would think so Because they have a lot of connectors And Salesforce is a partner of Google For instance, data transfer services to BigQuery So I would expect something, but I don't know Yeah, I don't know that, sorry I don't know the details of that So how is it different from data prep? From data prep That's a good question So the question was What's the difference to data prep? Data prep, as the name indicates Is really more for data augmentation In the beginning of the process So you're cleaning up your data You want to do exploratory data analysis Those kind of tools So data prep works with data fusion It would basically be a step of data fusion Data fusion is one level higher Here you can say really All the way from the beginning I want to ingest my files As I know, I'm just making something up You're getting mainframe files And you have to split, you know Do you have some abscidic logic Get that into CSVs Get that into BigQuery Do transformations inside BigQuery And then overnight You're ingested somewhere else That's something you would do in data fusion You can't do that in data prep So it's way more complex, yeah Then we announced Microsoft SQL server As part of Cloud SQL Now in beta That's pretty awesome So I managed a SQL server That's pretty great Especially if you then also use The managed Active Directory services Then we have BigQuery BI engine BigQuery BI engine I'm going to talk about later as well So I'm skipping it a little bit The one liner is This is basically in-memory BigQuery So BigQuery has a At the moment BigQuery is optimized for very large queries So the larger the query The faster you are basically But what happens if you really Just want to request one or two rows Very, very small What do you do? And that's where BI engine comes in So BI engine gives you Super fast access to your recent data It's basically like an in-memory caching layer That returns you millisecond latency For those kind of queries So it's really great And then last but not least We have Cloud Data Catalog And Composer are both now in beta And in NGA Data Catalog is our meta data store So that's where you find All of your different information At some point If you reach the right size on Google Cloud You don't really know anymore Who has your tables And who has your data sets And who has what And where they belong And Data Catalog gives you A unified view of all of those And finally AI and ML We have BigQuery ML If you haven't played with it Do it And we have BYOM Bring your own model That's pretty cool So that's a new BigQuery feature That you can build your own TensorFlow model And plug it into BigQuery And then call it as part of an SQL That's pretty nice So I'm not sure if Reza Is going to show a demo on that later Are you Reza? No, I have a small piece He has a small piece on it So definitely stay for Reza He'll talk about it later And he's going to talk about this Then we have new models for AutoML AutoML is our automated Machine learning solution Where we optimize the machine learning architecture For you even So you don't even have to choose The architecture of your model You're automatically You just have to give your source data And we figure out the best model One of the data scientists The machine learning experts on my team Said by now it's so good It's really hard to beat So even for machine learning experts To reach that level of AutoML quality Is serious work right now You can't just go into Like a Jupyter Notebook And define a quick model Like this is actually getting Really, really good And then we have a few new APIs Just to quickly recap AutoML is used if you want to Train your own model And then you have your complete Own model APIs you pay by call So this is like official APIs That we offer you Vision API for instance Where you just pay for Each kind of call So super easy Serverless even from You can just do it Now we have a lot more APIs So for instance We have a document understanding API That gives you sentiment analysis Structure, paragraphs All of those kind of things You have a product search API So now you can find your own products Or how people are reacting To your products We have the contact center That's for call centers So in call centers To understand what do people want Actually during a call And we have a recommendation engine As well So where we can Where we can give you recommendations For existing data And last but not least We have the AI platform The AI platform is our New umbrella term For all of the AI products Why do I mention it here It wasn't actually a cloud next Announcement All of you realize that Right, yeah I snuck that in Why? Because many people were Asking us in the past What's happening with Datalab And how does it work together With Scolab Which are our two Notebook solutions Now we have notebooks In AI platform And that's really The next step So you can just really Maintain notebooks And collaborate on them As part of AI platforms That's really nice ML remains separate Right? ML remains separate It's all part of AI platform now So what used to be Cloud ML is now Just called a job Or a task And then they run So it's all part of AI platform And it's now all together Because what you can do Now for instance Is you have a notebook That you're running In AI platform notebooks And then you want to train In Cloud ML engine And you can do that Directly out of the notebook You don't have to copy The code over somewhere And then submit it As a job into Cloud ML engine Cool. With that That was a very, very, very Quick run Through all of my Favorites So let me quickly check I'm half an hour in If that's correct So I don't see Okay, doesn't matter So, yeah I have a few more details Now in videos But before that I'm going to ask you Again for questions Because videos You know all of the Afternoon We're waiting for pizza So I want to give you A moment To quickly ask questions Any questions from Those people over there You have asked so many Questions already And I know you Have my email address So you can ask me later Any questions here To anything I said Any particularly Interesting announcement That you like Yes One question If you could elaborate A bit more On the superiority Of the data fusion Over the data prep Oh, okay Okay, so a little bit More difference Regarding data fusion And data prep Basically Data fusion Is based on an Open source framework Called CDAP And it's Creating spark jobs For you So the advantage Is again That you can Actually handle this You created In a nice UI But in the end It's a regular spark job So if you're coming From a spark world You feel comfortable Data prep is really One step back That's for data scientists Who want to clean up the data Who want to Who want to Apply a little bit Of business logic But this is all in the UI And you have no idea What's happening behind the scenes Right, this is really This is a magic tool So think of it As almost like serverless So this is a tool For data scientists Data prep That you use to clean up Your data Whereas data fusion Is a full blown UI Where you can Build very complex ETL pipelines And then they run On spark And you can investigate With all of the spark And I do tools That you know So what's the purpose of Data support? Data flow Well, data flow Is one of the runners That fusion supports So data flow Is similar to spark Right, so you can In data fusion You can say I want to run this on spark Or I want to run this On data flow I don't have a nice picture I will show you a nice picture But I didn't have A nice picture in this deck There's a nice picture On the website Okay, a few of my Personal favorites That didn't get enough call outs My absolute number one Is open source partnerships Has anyone heard of that? Because it wasn't a keynote But somehow no one Ever asked me about this So basically We announced That we now want to work Way closer than just a Marketplace with open source companies And that means That through Google Cloud You can now get the products Directly Like for instance Elastic search Including the billing And the support So you don't have An extra contract You don't have A separate support contact It's all completely Fronted by Google But we work with those companies We are not stealing The technology And just running it on premise This is still Completely this company So if you have Your own licenses You can bring them in So you have All of the freedom You still interact With those companies But through a single Paint of glass That's provided in the Cloud console And I think that's really cool So we have this nice Well So let me see if that If you can hear it If not we skip it But recently The open source community Has found That cloud providers Are not partnering with them But attempting to take away Their ability to monetize Open source We as Google Do not believe That that is good for customers For the developer community Or for software innovation As a result We have partnered With leading open source companies To deliver open source To our developer community And customers In new ways We're very pleased today To announce the first Integrated open source ecosystem What this allows you to do As a developer Or as a customer Is to use the best Of breed Open source technology You can procure them With standard Google cloud credits You get a single console A single bill From Google We support these along with the partners And as you grow And use these technologies We share our success With our partners So we're fostering This new open source community To ensure that open source Continues to thrive And has a vehicle To grow in this world Let's hear now From the chief executive officers Of some of our important partners About why this matters so much To the world To the developer community And to the ecosystem When you change infrastructure It's very much a kid To changing the infrastructure In a city It's no small undertaking When there's already buildings That exist That's why we turned To Google cloud as a partner When I talk to customers The common theme I hear from every customer Is they want to be able To innovate more quickly But the data landscape Is more complicated Than it used to be People need to actually bridge Between all these different systems All the devices All Okay Thank you So I'm not going to show You the full video But I think you get the idea And I think it's actually Really exciting Because sometimes it's just better That you work together with partners And I think it's a great solution We had this in the past Already with the marketplace But I actually prefer this way Because now you can Walk together with a partner Into a customer meeting And really make sure That you get the best thing Done for the partner And not have some product That doesn't Is not really offering All the things that the Partner wants to do And that was always A bit of an awkward situation And now we can really Work together with our Partners Anthos We spoke about it A little bit before It's our hybrid cloud Services platform The whole GKE on premise Is really only a very, very Small part of it It's the whole control plane Of all of your services So you can imagine it It's similar to for instance Like OpenShift Or PivotalCoutFoundly Those kind of things But on a way lower level Because we're not enforcing Any frameworks on you It's all working together With containers or VMs It's working on your Network stacks So it gives you All of the advantages Of a real service mesh Of a service platform Without locking you Into specific programming Languages Or specific frameworks Or specific build processes All of that is just taken away So it's just all open source Components You can configure them Anywhere you like And I mentioned for instance Before a traffic director This is then a separate product You can use it or not Cloud run as well So it's really a plug and play For an open service mesh platform What I wanted to show you Quickly is a really cool Tool that's called Anthos Migrate That migrates VMs From on premises Directly into the cloud As containers running on GKE So I'm quickly going to Run that video as well So what if I told you That we have a streamlined way Of taking this application And moving it to GKE Moving it into containers So that you don't have to Have the separate OS And VM dependency And maintenance Well that would be Kind of a game changer So what I'm going to do Is demo that We have actually built A vSphere plugin That's running here That's going to show you Over here All of the changes That will take place So we'll look at recent tasks To see what's happening And this plugin is also Connecting to our cloud Over a fan optimized connection So that all I need to do Is to go to GKE And from there I can initiate this migration So let me do that I'm going to go to My cluster This is a cluster Running in Europe And I'm going to go ahead And using our handy dandy Cloud shell Start a command line Environment And establish a connection To that cluster This is just standard So now I'm connected To that cluster And in that cluster I'm going to begin This migration So that's as simple as Cube, cuddle, apply From this file Which is A migrate animal file And what this file Is going to do Is it's going to initiate A shutdown of the VMs On-prem It's going to gracefully Shut those down Take a snapshot And then move that VM To this cluster over here And so let's go back To the cluster for a second Minimize this If I look at workloads You'll start to see That there are two new pods That are starting These are actually Stateful pods That are going to house The application It migrates over And that's going to Take a little bit of time So let's go back And actually see What's happening in the vSphere Screen I told you that We could monitor What's going on here And so you see here Hopefully you can see It's still even from the back You know the We have initiated Guest OS shutdown And we're re... So you can believe me It's going to continue Pretty cool But yeah You can watch it yourself If you like it Actually it works Believe me It was on the On stage as well So that's a That's a very, very cool feature Of Anthos We spoke about Cloud Run Already a little bit So I'm going to Not go so much into details Here But what I mentioned earlier Was Write the code your way Don't Don't rely on Any kind of Like serverless frameworks Just deploy wherever you like Deploy it as a cloud function Deploy it directly Out of IntelliJ Directly out of VS Code Or deploy it to Your cluster Or GKE on prem You can Directly use All of the serverless Serverless services That you're used to All of the other managed Services of course In Google Cloud Platform Pub, Sub, BigQuery Whatever you like You just connect to those And it's all In the exact same experience And it's all open source So there's no magic Behind that So that was for me Really cool Playing with that Was really nice And to be honest A little side note here Because I was also Like I'm an engineer Inside Google And for me It's really the point Where Google Cloud feels to feel Like we develop code internally It's really this Like we don't have to Sometimes client ask me Oh, but how do you run this application Or how do you run that And my answer is always Well, it's all services It's all containers It's all in bulk Like we don't have Those distinctions To be honest Like very often We don't even think About something as a bet Sure as a stream Like it's all It's just code And with Cloud Run We're really now coming To this point And that's For me really exciting I struggled a little bit When I was working with clients To make sure That I'm explaining it In the words That work in their context And not in mine And now we're suddenly Coming to a world Where I can just Give my experience That I have from every day Developing code Directly to the client And explain it In the terminology of Cloud Run Which I really like I don't have to Now struggle and talk about Ah, yeah, so You first do Jenkins And then It's a bit clunky For me to remember that And now I can finally Use that So that's really exciting For me personally To give all of the experiences From Borg That we already brought Into Kubernetes Now across the whole experience Not sure how many of you Have read the SRE book For instance But all of those principles Are part of that And go into the monitoring And the testing And releasing And canaries And all of those features Are part of it And it's very nice Because now you can talk About those things Rather than just having A big SRE book And Kubernetes And saying This is how it should Be done And now go Yeah, so that's really nice For me. Okay, I think I'm almost out of time So I quickly look around Okay? Okay, cool So I do I have a question Yes? I understood that actually We can send it Some web notifications Using Google Cloud Events Yeah And trigger it So if someone tried To ask like How would I decide Whether Cloud functions Or Cloud Run is better For me? Okay, so And the question was How do I decide Between Cloud Functions And Cloud Run Look, my personal opinion is And this is not Official Google statement That probably at some point You don't have to Decide that anymore Because it's just part Of your process inside Cloud Run Like it wouldn't be Very different products Anymore It's just a part Like you as a coder You really shouldn't care Like maybe there's Someone still in operation Success Or for cost reasons I want to optimize this For cost per request As opposed to cost Per instance Or cost per I don't know Of time Like those kind of decisions Someone will make Whether you are Making that decision I guess at some point Doesn't really matter anymore I would say So at the moment You can still use Cloud Functions But really give Cloud Run a try I think In the long term If you stay close To that model It's probably going To be It's probably Going to be better For you in the long Term Because you're also Yeah, you're not locked Into any other kind Of framework So you can always use Cloud Run You're not dependent On Cloud Run If you don't like Cloud Run Then you just take the code And copy it back into a function But at least you've tried Something that can do More than just A Cloud Function That's That's how I would argue Okay Cool Data from Victory Reza is going to give An amazing talk about that Later So I'm just going To mention it Huh? I'm not selling anything People are here only For you I'm like One of those bands That is like Before the actual Band That's That's why I'm here Yeah So For the ones who've seen Me before I'm a big BigQuery fan I do a lot of big queries So I'm just going through This again We spoke about the BI Engine integration So basically Our in-memory BigQuery store To provide that I have a quick video On this We also have A lot more templates Now in Dataflow So if you never did That in the UI Go to the UI Check out What we call the template So for instance Importing something Is much easier So test those out And for BigQuery itself That's a launch That went a little bit Below the radar But actually what's Super important Is something called The high throughput storage API That's the third way Of ingesting And reading from BigQuery So we only used to Have basically two ways Streaming our batch loads And for reading It was basically Running a query Or using the rest API And this is now A GRPC-driven API That for some customers Is like A bit faster Than anything they had before Because they're basically Streaming out the events Directly from BigQuery Without having to Translate it into JSON Or something like that And that's where we get A real big speed up So have a look at this This is actually My last video I'm quickly going to show this This is from Nikunz From AirAsia We moved it over to The BigQuery engine And a pipeline that we have We're able to go ahead And distribute this Daily And customers Or our end clients Can go ahead And drill down and use that So I'll go To the date range But if I was interested In Let's go ahead And say My nationality I can just click on that It's using the BI engine The background And filters everything Pretty quickly And if I was interested In this Float layer Or there all around I can go ahead And click on that And in sub seconds Goes ahead And changes The data across Which is Fantastic Because when You're in a meeting You can refer to this Dashboard Or anything So you can get up to Many, many users This is a copy Of the real one So obviously There's only one person On it right now But when we're in meetings You can get up to A lot of people on it We have around A thousand Folks in marketing And commercial So you can imagine We can get about 200 people Looking at this On a daily basis Not all concurrently But concurrently I usually see Around 50 to 60 People using it And now with BI engine The way that it works It really makes it easy And fast For people to go ask That question And leave the meeting With real concrete Action items Versus actually Waiting for a report Back and asking A question And then having another Meeting on a report So it works quite well Let me just Reset everything And you can see here As I filter along It's pretty responsive And fast And how it works So it's quite impressive For us So yeah So that was BI engine Yeah, really cool I think Especially because it's live And just to Give you an idea Sorry I can't get it The full screen I have to go first And this is like On all of their sales data Right? This is not like Some small data set We're talking about basically Just like It's the full sales database That's a big query Right? Massive scale You don't need to create Views or materials Views or whatever Status anymore This is just On all of your data Because that's what And so he actually Shows it in the demo Last but not least Security enhancements We have a lot Of security enhancements I mentioned in the very Beginning Have a look at Security command center It's pretty awesome Also have a look at Fosetti Fosetti is our security Scanner That's actually developed By professional services To Not only scan Google Cloud But scan Anything else as well So it's An open source solution Have a look at that For rules And one last thing That I really want To tell you That we have on Google Cloud And that right now Clients have asked me a lot Is two new features That we have That's access approval And access transparency Services It's interesting That this is never Really mentioned So publicly So what does it mean? Access transparency Means you see In your audit logs Every time Anyone from Google Accesses anything From you That no other Cloud provider Offers that to you You see every time If a support engineer If a VM is migrated Or something happens A security patch is applied Whatever This always shows up In your logs And access approval On top of that Is now a new offering That we have That you can actually Decide whether you want That access to happen Yes or no So that's really Really cool That gives you Complete full control There's no backdoors Nothing This is complete control Over everything you run On Google Cloud So that's Actually a really Really cool feature And you see Everything in your audit logs Everything that's happening Even for infrastructure Components that you Normally don't know And that you Don't really care about But sometimes you do And that's It's still important So what's the definition Of the unit What is the access For the access Yeah, so the question Is what is What's the unit So there's Very different ones That can be anything From a VM security patch If you have Auto-upgrade enabled Down to for instance Let's say You're raising a ticket To cloud support And saying My data flow job Is not running anymore And I don't know why Then that person Might have to access The logs of that job You will see that The job access Let's say You raise a ticket To cloud support And say I'm running this BigQuery query And it's taking too long And then that person Also runs the same query Against your original Dataset That will also turn up There So everything Is very fine Great Not just like No It goes down To every single action That we're doing On your behalf And that's it Actually for me I think I'm pretty much On time Did I go a lot Over time Okay cool Awesome So yeah Thank you all very much Any last questions Otherwise I'm still here Okay cool Thank you very much