5 architectural innovations in VMware Cloud on AWS that you can’t miss

Let's begin. As you all know, today's world is a digital first world and around 95% of the CEOs see that they want to move from their physical equivalent to the digital equivalent. But doing this digital transformation is not very easy.

We have seen that customers have been using the data center, but they have been facing multiple challenges there because of static infrastructure, manual operations and waterfall development processes. And that's why over the last few years, they have started migrating their workloads to the cloud because public cloud inherently is more elastic, more automated, more agile and it helps them migrate and modernize their applications faster.

But when we talk to our customers, they always mention that this is not as easy as they thought before starting their migration journey. And there are multiple reasons for that:

First is different skill sets and processes. So when they migrate to the cloud, they have to learn new tools, new skill sets, there is a big learning curve there. They have to hire some new staff as well.

Also operational inconsistencies. So they cannot use the same tools and processes they have been using in their on-prem infrastructure. So there are divergent tools that they have to manage when they move to the cloud, different tools and management consoles.

I already mentioned that also different machine formats. So when they move to the cloud, they have to re-architect or refactor their applications. And that is a very cumbersome task.

Also incompatible configurations and capabilities. And because of that, it adds a lot of complexities and it delays their migration process.

So we figured out that customers when they start their migration journey, there are three key areas where they feel it is very challenging to migrate to the cloud:

  1. While migrating to the cloud. These are some of the key challenges when they start migrating.

  2. While operating in the public cloud.

  3. While modernizing their applications.

First thing is when they migrate to the cloud, the cloud is designed with high granularity of services. So if they try to modernize any application, they have to assemble dozens of different services and these services are highly independent of each other. So they have to figure out how am I going to integrate this service? Am I going to do it myself or am I going to outsource it to some other company? So a lot of different challenges involved when they start operating in the public cloud.

Then the next set of challenges is while modernizing their applications. So this one size fits all strategy doesn't work the way it seems. Because the value delivered by the applications or their infrastructure needs, their devops processes, architectural designs, they are all different for different types of applications. So they cannot do the modernization at once. There needs to be a way to do the modernization pragmatically step by step.

And because of all these challenges, customers have seen that there is a reduced value when they migrate to the cloud. Also, it takes longer time for them to migrate to the cloud. Because of all these challenges, also higher risk is involved. A lot of security concerns that customers have raised and last very important one is higher costs. So they see that after moving to the cloud, actually their costs have been increased compared to their on-premises environment.

So in order to address these challenges and help customers accelerate their cloud transformation journey, we started this service, VMware Cloud on AWS, around six years back. So VMware and AWS, two powerhouses in private cloud and public cloud respectively, came together and they started with VMware Cloud on AWS. And we have been doing a lot of new feature releases every quarter.

So we launched the GA of this service in 2017. And since then, every quarter, we are releasing new features and capabilities for this service. And Oleg will cover some of the latest features in the later section.

Let's look at the service itself. On the left hand side, you can see that this is your on-premises infrastructure, on the right hand side, it is the AWS cloud and in the middle, you can see this is VMware Cloud on AWS environment. So what we are providing to you is the entire VMware software defined data center stack that includes vSphere, NSX along with vCenter management. And that is running on bare metal AWS E2 instances. So the key point here is this is not a nested virtualization. So everything is running on bare metal E2 instances. So there are no latency issues or anything.

And this service, this is the only service that is delivered, operated and supported by VMware. So VMware is your single point of contact for any kind of support. VMware is responsible for providing all the updates, upgrades, patches. So you don't have to worry about anything in terms of lifecycle management.

Also, you can purchase this service on-demand model where you can buy the service on hourly basis or you can go for flexible consumption options, like you can pay monthly, you can pay like one year or three year subscription. So there are a lot of different consumption options available there.

Then operational consistency with the on-premises environment. So everything is based on the same vSphere, vCenter, APIs that you have been using in your on-prem environment for the last 20 plus years. So a lot of operational consistency and that helps you in doing your migration.

Next very important point is seamless large scale workload portability. So we are providing something called VMware HCX and it is part of the service. So you are not paying anything extra for that. But with VMware HCX, you can do bulk migration. So you can literally migrate hundreds of thousands of workloads at a time. Also with vMotion, you can do live migration. So there won't be any downtime while migrating your workloads.

Another key thing is hybrid operations. So we are providing some hybrid capabilities like for example, there is something called Hybrid Link Mode. So once that is enabled, you can have a single pane of glass view for all of your resources that are there in your on-prem environment and in VMware Cloud on AWS.

Next very important point is global AWS reach footprint and availability. So today the service is available in 26 different regions and every quarter, we are extending the service availability in different regions and I'll show the regional availability map in a bit as well.

Next very important point - direct access to native AWS services. So we are providing you a special engineered elastic network interface connection. So using that you can access around 200 plus native AWS services from your VMware Cloud on AWS environment. So that is especially useful if you are trying to modernize your applications using any of the AWS services.

And last very important one is suppose you want to scale your storage capacity independent of compute, then you can do that as well. So we are providing multiple options for that and we are going to talk about that with our latest announcements as well.

These are some of the key numbers:

  • Service is available in 26 regions and expanding every quarter
  • There are 25 plus global certifications available on the service today
  • To set up an SDDC, it takes around 65 minutes
  • According to your requirements, you can add hosts in around 12 minutes
  • Today there are 300 plus ISV ecosystem partner solutions that are validated on VMware Cloud on AWS
  • 99.99% SLA with the Stretched Cluster capability
  • Around 650 plus Solution Competent partners

Let's look at some of the key differentiators of the service:

  • What we are providing is a complete enterprise ready turnkey solution, a complete package. So you can just have that and you can deliver enterprise applications.

  • Fast and non-disruptive migration. So you don't have to re-architect or refactor your applications and the migration process is very smooth and very fast compared to any other public cloud options.

  • Pragmatic and non-disruptive modernization. You can modernize applications pragmatically according to availability of your resources, your timelines.

  • Compelling TCO. I'm going to show the TCO slide in a bit as well, but we are providing some resource optimization options. Also in terms of migration you are saving a lot of costs there. So you can reduce your TCO compared to your on-prem investment or other native cloud options.

  • Consistent and familiar cloud platform. Whatever you are using today, in terms of tools, skill sets, processes, you can use the same ones in the cloud with VMware Cloud on AWS. That is a key differentiator.

I already talked about it - this service is delivered and supported by VMware and its partners. All updates, upgrades, patches provided by VMware. This is a mature service - operations and response teams are also in place for any kind of support.

Let's quickly look at some of the key use cases. The first and very widely adopted use case is application migration and modernization.

So we have seen customers using VMware Cloud and AWS for their scaling of line of business applications. Or there are some customers who have some application needs where these applications have some architectural requirements and they cannot move these applications to any other cloud, rather they have to use VMware Cloud and AWS for those specific applications that have specific architecture needs.

Also for the next generation applications - if you want to enrich the value of your existing applications, and you want to modernize those applications, you can use VMware Cloud and AWS.

Next, this is one of the most widely adopted use cases as well, which is data center migration. So customers have been using VMware Cloud and AWS for outsourcing their data center. They are not happy with their colocation facility or their lease is going to expire. They are not happy with the SLAs so they want to get rid of their data center and move completely to the cloud. They have been using VMware Cloud and AWS for that. Or say there is an infrastructure refresh cycle coming up and they want to modernize their infrastructure - they have moved their workloads to VMware Cloud and AWS.

And then another widely adopted use case, especially we saw recently, is disaster recovery. So customers have been using VMware Cloud and AWS for ransomware recovery needs. Or if they want to move their DR setup out and move to DR as a service in the cloud, they have used VMware Cloud and AWS. And we have our in-house solutions for that like VMware Site Recovery or VMware Cloud Disaster Recovery. So those work as add-on solutions on top of VMware Cloud on AWS - you have to purchase those separately. Or if you are using any of the partner solutions, then we have a lot of partner solutions that are certified on VMware Cloud on AWS that can help you with your DR needs.

And then the last one is data center extension. This is the second most widely adopted use case. Some of the key scenarios here are - customers want to expand their footprint across different regions. Or they have some on-demand capacity needs for variety of reasons, like seasonal spikes in demand or some event coming up and they want to set up IT infrastructure for that event and tear it down once the event is over. Or if they have some virtual desktop needs in the cloud, like universities or lab setups. So for that, they have used VMware Cloud and AWS.

This is the packaging slide. I have mentioned most of the things here but basically you can purchase this service on a per-host basis, and when you purchase the host what you get is the AWS hardware, the two hardware instances, you get the VMware SDDC software, you get support included, lifecycle management is included. And there are some optional add-ons that you can purchase. You can purchase this from VMware or AWS, or any of our partners. And there are multiple payment options available. Also there are different consumption options available as well.

For VMware Cloud on AWS, we are offering two key instance types. First is i3en metal - these are some of the key specifications of i3en metal. This is more useful for storage bound workloads. And then the second one is i4en metal - so we announced this last year. This is for general purpose workloads and you can see all the specifications here.

We made another announcement for another instance type today morning - we will cover that later.

This is our regional availability map - the service is available across 26 regions. Asia Pacific Hyderabad is the latest region that went live last month and we have two more regions coming up.

These are our compliance certifications - you can see we have global compliance certifications from different regions.

Now let's talk about TCO. We did some studies with the help of analyst firms. On the left hand side, this is the study we did with Forrester where they interviewed some of our existing VMware Cloud and AWS customers - these were customers who were using traditional vSphere on-premises. What they found after interviewing these customers was they were seeing around 53% recurring savings in terms of infrastructure and operations costs when they moved their environment from on-premises to VMware Cloud and AWS.

On the right hand side, those are interesting numbers as well. So this is the study we did with IDC. They also interviewed our existing VMware Cloud and AWS customers who were using some other native public cloud options before moving to VMware Cloud and AWS. And what they saw after moving to VMware Cloud and AWS is their costs were reduced by more than 57%. They saw more efficiency and also there was less unplanned downtime, around 83% less unplanned downtime. So those were some key stats we found after interviewing existing customers.

When we talk about this TCO - we want to do an apples to apples comparison - what are some of the key elements of this TCO comparison? Because as I mentioned, we are providing the entire package of vSphere, vSAN, NSX - you are getting everything together. With this entire package we are providing, customers are already saving around 27% of costs.

Then, as I mentioned, we are providing resource optimization techniques like CPU oversubscription, memory overcommit, VM optimization. With this customers are able to increase savings to around 34%.

Then, we are providing additional features in terms of enterprise compute, networking, security, and storage. With these additional features, cost savings increase to around 38%.

And lastly, because you don't have to re-architect or refactor your applications, you are saving labor hours and time. With that, your savings increase to more than 50%.

Next is our customer logos - this is not an exhaustive list, these are only customers who gave us permission to use their logos. But you can see we have customers across different geographies and industries.

Now let's talk about a couple customer stories as well.

Kingston University, a renowned institute in the UK, wanted flexible and agile infrastructure and their IT team wanted to support the university's transition to a digital learning environment. So what they did was use VMware Cloud on AWS and migrated around 300 VMs in less than 14 months, which was a very fast migration.

Quality Bicycle Products is another retail and wholesale business in the Americas. They wanted to consolidate their data centers and move completely to the cloud. They used VMware Cloud on AWS and were able to reduce their data center footprint by 90% and literally migrated 200 VMs without any downtime in just 3 weeks.

And then another use case and customer story - this is a ransomware recovery use case. vGIS, a financial services company in the United States, recently faced a ransomware attack and wanted a robust solution to deal with these attacks. They used the Ransomware Recovery add-on on top of VMware Cloud and AWS and were able to find a robust solution for these ransomware attacks. This is a very recent customer story.

As I mentioned, there are these add-on services on top of VMware Cloud and AWS for different needs like disaster recovery, cloud management, advanced networking, or if you are interested in using containerized applications, there is VMware Tanzu. All of these are available as add-ons on top of VMware Cloud and AWS.

We have a growing partner ecosystem - 650+ solution competent partners and this number grows every quarter. I already talked about the 100+ ISV ecosystem partners and around 300+ ISV solutions validated on VMware Cloud on AWS.

We are running a trial environment for VMware Cloud on AWS. If you are interested in the service but want to try it out first to understand how it works and whether it is suitable, you can scan the barcode and go to the trial page and sign up for a free 1-month trial. You get 1 host of i3en metal or i4en metal and you can try it out before moving to the production environment.

We are also running a Multi-Cloud Adoption Program - with this program we provide credits to help with your migration efforts. You can work with partners and VMware to start the migration journey. Check with your VMware rep for this program.

Now let's talk about some key announcements for VMware Cloud on AWS.

The first one we announced today is disaggregated compute and storage with the new M7i instance type. This is a groundbreaking innovation for VMware Cloud on AWS.

First let's talk about why customers need disaggregated compute and storage. Today, customer needs are changing - they want to support modern AI/ML workloads, more flexibility, better performance for next-gen workloads. Also they want to reduce costs as customers are very cost-sensitive.

Previously with i4en and i3en you got compute and storage tied together. When you purchased a host you got compute, storage, networking together. But now with the M7i instance type, we are disaggregating storage and compute. This gives more flexibility in choosing compute and storage.

Here are some key specs on M7i - it is powered by 4th gen Intel Xeon Scalable processors, codename Sapphire Rapids, with turbo frequency up to 3.9GHz. With this processor you get accelerators that are important for accelerating matrix operations and AI/ML workloads. This is a very high performance processor.

You get 48 physical cores, 96 logical cores with hyperthreading enabled, up to 37.5Gbps networking speed, and 384GB memory.

And the key advantage over here is you have the flexibility in terms of choosing your storage. So you are not getting vSAN as a default storage along with when you purchase the host. So you can go for Amazon FSx with a TAP on TAP, which is one external storage that you can use or there is another option of VMware Cloud Flex storage that you can attach to your cluster.

Um so let's talk about some of the key benefits over here from the customer's point of view. Um the first one as I mentioned is flexibility. So now you get this flexibility to choose your compute to choose you choose your storage. So you have multiple instance types like i3 and metal i4 i metal or, or m7i. So according to your use case needs, you can choose your compute. Then depending upon your storage needs, you can choose either vSAN or you can go for Amazon FSx or you can go for um VMware Cloud Flex storage. So lot of flexibility and options, scalability.

So scalability say suppose you want to scale your storage independent of compute. So previously, if you have to scale your storage, you have to purchase host and then unnecessarily you are paying for the compute, right? But now with this, if you just have to scale storage, you can just purchase storage because compute and storage they are independent.

Now, next one is cost efficiency. So with all this scalability and flexibility, you have the option to choose how much storage you want, how much compute you want and you are not paying for unused or redundant capacity and there itself you are saving lost a lot of costs and last one is improved performance. So as I mentioned, this is running on the latest Intel's fourth generation processor, it is very high performance processor. So you can run your AI/ML workload. So that is the biggest advantage for for especially for the AI/ML workloads.

And these are some of the key use cases over here. Um so the first one is some CPU intensive workloads and some of the examples here is like image processing workloads or video encoding or gaming servers. Um then another very important one is AI/ML workload. So say if you have any natural language processing workloads or code generation or recommendation systems, you can use that. Um then the next one is workloads with limited resource requirements. So as I mentioned, if you need like less amount of compute or less amount of storage and you want to decide what you want to use, then you can go for this instance type versus the others where you need limited amount of resources. And last one is um ransomware or disaster recovery. So say if you want to have a pilot light solution for your DR needs or if you want to have a ransomware recovery set up, then you can go for this instance type and um then I'll hand it over to Oleg.

So Oleg will show you a demonstration for m7i instance type and then he'll take it further with all the other announcements. Yeah, thank you very much. So it's really impressive what can be produced, right? And what we're offering right now is a service. But with that, let us take a look at some of the technical details, right?

So let us start first with m7i in order to provision the new instance, it's actually nothing changed, right? You will use the same cloud services portal by logging to cloud services portal as starting the SDDC deployment visit, right? Obviously, we need to provide the name, we need to select the region um like nothing changed, right? We do support with m7i multi host deployment with a single AZ at the moment and with the instance type, right, you see this new m7i being now available as a part of this visit.

And that's basically it, right. So we, we will take care of provisioning management, data store, right? Because it's a desegregated instance, right? It's no storage. So we need to map a storage that is something that will be done by, by us, right? And we will deploy SDDC exactly the same, right, connected to AWS account, everything all in but just not having a VISA.

And once we log into SDDC after the provision as a part of the storage tab, that will be on you to provision the workload data store, right? So where you gonna run your VMs, it will be customer responsibility to add an external NFS data store. It could be, as Tonali mentioned, it could be based either on VMware Flex storage or it could be based on FSx for net upon tab, right. So that's the key here just to understand the difference in the deployment process.

Also what is important to mention in us SDDC, you can deploy FD DC as shown based on m7i right, only m7i clusters, you can mix and provision first cluster using c7i and seventh cluster using other different host type, right? So basically what does it mean you can provision your management cluster using use classes instances, right? Not paying for the full i5 host or you can add this class or desegregated storage instances to your existing SDDC while on 124 right? So those are supported use cases for m7i deployment at the time of the launch with that, let me switch back and we are happy to take your questions right, regarding this new instant type after the session.

But now let me switch to a number of new technical details and exciting things, right? I have only what 20 minutes, 25 minutes left to go over. Like, you know, a lot of things that we provision to the service for the last one year, the one he added at no additional cost, our management software stack, a operation, a automation and area for locks. Meaning right now you have access to full lock solution to full monitoring solution and automation as a part of the same service, which is a great value at no additional cost.

We do have advanced load balance available as a service. Uh also attached to VMware cloud on AWS if you decide you need for advanced load balancer and luck, but not the least, right? We just recently released a new SDDC version, right? As sole mentioned what we are doing, we are deploying KXinsx and vSAN if needed on a if VA is a part of the store stack, right? And with that, we always staying on top with the current vSphere version, right? With the new one and 24 we added support for vSphere update two, which is the latest one on available on the vSphere version. And that is in new release, all of the value that will come in new enhancements that I'm going to talk about and probably the important one, there is 124 the enable vSAN express torch architect.

That's really a new and changing factor for your uh instances with vSAN. It's give you another way to use local and v medi with much more resiliency, better, better space to better space usage and a better performance, right? So in terms of resiliency and what you provide with ESA is when a disc fail occurs, it would not take the whole dis group out. It will just affect an individual disk and it's really a great deal, right? Because when the whole dis group fails, we need to recreate all the data for this, this group, right? Which may be a lot of data on the hose that we for individual dis uh that will be much less data, much less time to synchronize much less network usage.

So vSAN ESA is now available upon request and it will be available for all SDDC running KOI in a in a near time. Um new thing, right? I hope that everyone in the room know cloud sizer. So cloud size is your primarily tool to size us SDDC environment, right? It's really the key to get the right sizing to get it done right from the beginning. And the new enhancement for cloud size is now the beacon size for vSAN ESA it also change as I mentioned capacity requirements, right? Moving on with a ransomware recovery service and a ransomware recovery is now one of the key things happening in the world, right?

We all know unfortunately all these attacks is a pretty common thing nowadays. So we enhanced our ransomware recovery offering with adding new services and new functional available as a part of V CDR and ransomware recovery. Let us now go through some of the important things through compute storage and networking just to recap all technical innovation we have with the service.

And I think this one elastic DRS is probably one of the key innovation we have is we work out on a W service, right? It's not that you are running on top of S cloud, it's our ability to leverage elastic capacity. And that is exactly what elastic DRS does, right? Is basically monitor your consumption of resources on us SDDC and can help you to scale up or scale down without human intervention, your service.

And before that, as you may remember, elastic DRS were limited to some predefined policies, right? And as in the where we, we love to give choices and right now we support custom S policies, meaning you can define your own threshold for scaling in and scaling out here SDDC helping to match and to match your workload consumption.

We enhance life cycle management. And I truly believe that the way how we work out on AWS dust life cycle management is the best. What is happening with that service? It's fully automated, it's taking care not only of GX I updates NSX update V and updates, we're also doing automated firmware updates. And even so the only thing how you would notice is that is looking at the logs, it's really, it's no influence on your workload because the way how AWS uh hardware is working, we were able to recycle just services without recycling the whole.

Yes, SX I host storey and I really would like you to pay attention to that. As you know, we launched a new feature and the ability to attach an external NFS data store one year ago. And in order to use this feature, uh you will need to go through uh transit connect everyone in this room, probably knows that transit connect is a great routing solution is a great thing to have. However, transit gateway is also associated with additional cost that you will need to pay for aggress traffic flow. Right? Obviously enough, it's good for networking use cases. However, for storage use cases, it might lead for additional cost.

So what we did we introduced now a VPC peering another way to attach your VPC where FSx is deployed to your VMware cloud on AWS SDDC, reducing cost on transmitting traffic. It's a number of requirements for VPC buring we now support only uh single AZ, right? With VTGV, you can do multi AZ, right? For obvious reason because for that's a different way how your networking traffic being crowded, right? So VPC pairing only works with single AZ uh and same region. However, it's really key if you are now using external NFS data store for FSx, consider migrating from transit gateway to VPC ping because that will save you on cost.

And not only that right for as we are investing now a lot as you can see on these desegregated instances, we are also looking to improve overall performance accessing can data store is that we now change the way how we are connecting to external NFS data store, either flex storage or FSx by improving the number of connection and enabling jumper frame on our management interface. And everyone here with extensive VMware. Yes x I background would know that management like management interface is used to access NFS data store, right? So enabling jumbo frames basically increase a number of packets and what we can like payload in a packet that you can send over a single moment in a time.

Let us also take a look for news for VMware Cloud flux storage. It's now available through AWS resale, which quite a good thing for all AWS customers, right? It's available in 17 region and we now expanding file system per region and making it compatible with VMware cloud disaster recovery, which actually might be a good use case allowing you to save on the storage consumption with disaster recovery solution.

Networking is always the key, right? So networking is the way how you consume application and workload running on your VMware cloud on AWS SDDC. This that we evolved from providing you access to network and security tab within your VMware cloud on AWS console. And now we have so many new features to offer for networking that we need to find out a new way to manage it.

So that's why we now moved away from exposing features inside of network and security tab to a full NSX manager. NSX manager is instantly available. When you log into us SDDC, you just click on NSX manager button and you will get access to NSX manager and network and security tab is now serving only management purpose that might be exchange for your data operation.

So make sure to review your operation work box, right? Because that's also helps you not only to improve the way how you manage networking but also a speed for SDDC council because right now, we don't need to wait for NSX manager, right? We just show you overall monitoring information on this tab.

Um IPv6 support. I know that it's a lot of us out there for full IPv6 support out on SDDC. We are working towards this goal. With enabling IPv6 traffic inside of SDDC, ability to create a segments with IPv6 and also ability to route IPv6 through direct connect, right? We still it's a number of items on the road map, like support for IPv x through uh management interfaces for v center, right. However, we are definitely on a target to work towards enabling full IPv6 support for SDDCs.

And that was a long, long waiting for our customers, right? The owner that it's it's the need to provide a way to just audit what is happening without being able to change settings. And before that in our especially NSX for networking, it was like on the read only or you can get a full admin control. It wasn't an f for customer. So with recent 124 release, we now enable four additional NSX role to help you more gran manage permission for accessing NSX manager, UI I'm sorry.

And now for migration, right? We all know that like eight out of 10 deployment for VMware cloud on AWS is basically to help you establish a hybrid connectivity and even like it's a number of questions. Ok. What if I would like to migrate my physical server or what if I would like to move for some reason, issue two instances that's now a new functionality of vCenter converter.

vCenter converter is available once again for your workload and can help you to migrate physical server or two instances directly into the workload on AWS and STX. So STX is our primarily tool to help you manage workloads across uh hybrid environments and it helps you to migrate workloads, right? And STX is now enhanced to support a PV6 as well.

Moving forward this golf cloud, right? Don't forget that we have also VMware cloud on AWS available in a key golf cloud region, right? So let me just recap that we now have we high compliance along with other key compliance certification for golf cloud. And we worked a lot and i think the main and the key things we brought into VMware cloud on AWS gov cloud. Nowadays, its support for VMware TX, right? It was a long outstanding task to help our government customers to enhance the migration capabilities. And this new release of gov cloud 122 we now have full compliance TX solution.

Moving further, we support VMware cloud on AWS outpost and for VMware cloud on AWS outpost, we just recently enabled multi rack and multi cluster support to help you scale your on premises deployment of VMware cloud on AWS with VMware cloud on AWS outpost. And with that, we also provide you a lot of tools to help you facilitate your hybrid cloud deployment.

And one of them is to go and help you understand what will be the best path to VMware hybrid cloud environment and then analyze your workload and then experience it by providing you access to a hands on lab or free trial programs. We have a bunch of new uh technical documentation available on our T zone in new hands on lab, we fully refresh our hands on lab offering. So please make sure to check our hands on lab and.

你可能感兴趣的:(aws,亚马逊云科技,科技,人工智能,re:Invent,2023,生成式AI,云服务)