Migrate and modernize your database applications (L300) | AWS Events
Summary
TLDRビデオスクリプトは、AWSのデータベースサービスの現代化について解説しています。Barry Morrisが主導するこのトピックでは、データがAIの基礎であり、クラウドへの移行がビジネスの機敏性、スケール、および可用性を高めると語られています。Amazon.comがリレーショナルデータベースからノンリレーショナルデータベースへの移行を通じて、検索、顧客フィードバック、推奨エンジンなどの機能を強化した経験が紹介されています。また、クラウドネイティブのデータベースサービスを使用することで、アプリケーションの柔軟性とスケーラビリティを高め、ビジネスのイノベーションサイクルに再投資することができると強調されています。
Takeaways
- 🌟 AWSのデータベースサービスは、非関係型データベースを含む多岐に渡るデータベースサービスを提供しており、クラウドへのデータベースの移行とモダン化の過程で重要な役割を果たしている。
- 🚀 Amazon.comは、従来のリレーショナルデータベース技術から、検索、キーバリュー、グラフ技術などの多様なデータベースに移行し、クラウド上で高いスケーラビリティを実現している。
- 🔄 クラウドデータベースの移行とモダン化は、企業のビジネスの敏捷性、スケール、および24/7の可用性を高めるために重要なプロセスである。
- 🛠 従来の3層アプリケーションアーキテクチャから、マイクロサービスやコンテナ化されたアプリケーションへの移行が、クラウドネイティブなアーキテクチャを実現している。
- 🔑 データベースの選択は、アプリケーションの要件に応じて最適なものを選ぶことが重要で、これによりパフォーマンスと柔軟性が得られる。
- 🔍 クラウドへの移行において、マルチテナントアーキテクチャを検討することは、コスト効率とリソースの最適化を可能にする。
- 📈 Amazon Auroraは、従来のMySQLやPostgreSQLに比べて大幅なパフォーマンス向上を提供し、ストレージの自動スケーラビリティを備えている。
- 🛑 Babble Fishは、SQL ServerからPostgreSQLへの移行を容易にし、ダウンタイムを最小限に抑えることができるサービスである。
- 📊 Amazon TimestreamやNeptuneなどの特別なデータベースは、特定のユースケースに対応し、ハイパースケーラビリティやリアルタイムのデータ処理を可能にしている。
- 🔌 AWSでは、データベース移行サービス(DMS)やスキーマ変換ツールなどのツールが提供され、データの移行と分析のためのETLプロセスを簡素化している。
- 📚 現代のアプリケーションでは、データの分析と機械学習への活用が重要で、AWSはこれらのプロセスをサポートする包括的なサービスを提供している。
Q & A
AWSのデータベースサービスのどの部分をBarry Morrisが担当していますか?
-Barry MorrisはAWSで非関係型データベースを含む多数のデータベースサービスを担当しています。
データベースの現代化とはどのような意味を持っていますか?
-データベースの現代化とはクラウド上でデータベースを最新の技術に合わせて更新し、パフォーマンスや可用性、スケーラビリティを高めることを指します。
Amazon.comはどのようにしてデータベースの限界に達しましたか?
-Amazon.comはリレーショナルデータベース技術をベースにいても、その限界に達し、その後にDynamo DBなどの新しいデータベース技術を導入してスケーラビリティを高めることを決めました。
AWSはどのようにしてデータベースの選択と使用を最適化していますか?
-AWSは適切なデータベースを適切な用途に使用することで、パフォーマンスとアジャイル性を確保しています。これにより、データベースの選択がビジネスのニーズに合わせて最適化されます。
マイクロサービスとは何で、なぜクラウド上でスケーラブルですか?
-マイクロサービスはアプリケーションを小さく、独立したサービスに分割することで、それぞれのサービスが独立してスケールできるアーキテクチャです。クラウド上で実行することで、リソースの柔軟な割り当てが可能でスケーラビリティが高まります。
AWSのデータベース移行サービスのデータコレクターとは何ですか?
-データコレクターはAWSのデータベース移行サービスの一環で、自動的にデータベースのインベントリを収集し、分析することで、移行の計画を容易にします。
Amazon Auroraは何で、どのような利点がありますか?
-Amazon Auroraは互換性のあるリレーショナルデータベースサービスで、MySQLやPostgreSQLと同様のSQLをサポートし、高性能とスケーラビリティを提供します。また、ストレージが自動的にスケーラブルであることが大きな利点です。
AWSではどのようにしてデータベースの多テナンシーを管理していますか?
-AWSでは、単一のアプリケーションイメージとデータベースインスタンスを共有する完全な多テナンシーから、各顧客に専用のインスタンスを提供するまでのスペクトルで管理しています。顧客のニーズに応じて適切なアプローチを選択します。
Amazon Dynamo DBはどのような種類のデータベースですか?
-Amazon Dynamo DBはキーバリューストア型のデータベースで、非常に高いスケーラビリティとパフォーマンスを持ち、大きなデータセットに対する高速アクセスに適しています。
AWSのデータベースサービスはどのようにしてビジネスの迅速なイノベーションサイクルに貢献していますか?
-AWSのデータベースサービスは、データベースの管理や保守の重労働を自動化し、開発者がビジネスロジックやイノベーションに集中できるようにすることで、迅速なイノベーションサイクルに貢献しています。
Outlines
😀 クラウドデータベースの移行とモダン化
Barry Morris氏はAWSのデータベースサービスのリーダーとして、データがAIの基礎であり、オペレーションデータベースがデータの基礎だと語りました。顧客の多くがクラウドに移行し、クラウドデータベースへの移行を経験しています。Amazon自身もリレーショナルデータベースの限界に達し、多くのデータベースサービスに移行しました。また、現代のデータベースアプリケーションには、 Agility、Scale、および24/7の可用性が求められます。AWSはモノリスではなく、マイクロサービスアーキテクチャに移行し、スケールとコスト管理の両面で利点を享受しています。
😉 クラウドへの移行とマルチテナンシーの課題
クラウドへの移行は、データセンターの終了やコスト削減、デジタルトランスフォーメーションなどが理由として挙げられます。クラウドへの移行を通じて、顧客はコストを削減し、イノベーションサイクルへの再投資を望んでいます。ISVにとっては、クラウドへのサービス提供に伴いマルチテナンシーの問題があります。単純なマルチテナンシーでは、小規模な顧客に対して過剰なプロビジョニングが行われ、コストがかさんでしまう一方で、完全なマルチテナンシーでは、顧客間でアプリケーションとデータベースを共有するため、複雑さが増します。
🎯 データベースの分析と移行戦略
データベースの移行は、分析から始まるプロセスです。顧客は新しいアプリケーションを構築するか、既存のアプリケーションを分析し、削減、統合、リフトアンドシフト、またはモダン化を選んで移行します。AWSは、データベース移行サービスのデータコレクターツールを提供し、手動での監査を自動化して、データソースのインベントリを作成します。これにより、顧客はどのデータベースから始めるかのインフォーメーションを得ることができます。
🛠 データベース移行ツールの活用
AWSのデータベース移行サービス(DMS)とスキーマ変換ツールは、移行プロセスを支援します。スキーマ変換ツールは、顧客が最適なターゲットデータベースを選択し、スキーマ変換やコードオブジェクト変換を支援します。DMSは、データの移動と最小限の停止時間またはほぼゼロの停止時間のためのデータレプリケーションを支援します。また、AWSでは、Aurora PostgresやAurora for MySQLなどのマネージドサービスを提供しており、顧客はこれにより、インフラの管理から解放され、イノベーションに注力できます。
📈 パフォーマンスとスケーラビリティを兼ね備えるデータベースサービス
AWSは、さまざまなスケーラビリティとパフォーマンスを提供するデータベースサービスを提供しています。Amazon Auroraは、従来のMySQLやPostgreSQLに比べて大幅な性能向上を提供し、ストレージの自動拡張性を備えています。また、DynamoDBはキーバリューストアとして、膨大なスケールでの要求に対応できる能力を有しています。DocumentDBは、ドキュメントデータベースとして、MongoDBと互換性があり、スキーマの柔軟性を提供します。
🔄 データのリアルタイム分析と移行
リアルタイムデータ分析と移行は、ビジネスにとって重要です。AWSでは、ストリーミングデータのためのKinesis、MSK、バッチ処理のためのEMR、Glueなどのサービスを提供しています。これにより、顧客はデータパイプラインを構築し、データの移動と分析を効率化できます。また、Amazon QuickSightなどのビジネスインテリジェンスツールや、Amazon SageMakerのような機械学習サービスを活用して、データから洞察を得ることができます。
🚀 ベクターデータストアと生成的AIアプリケーション
生成的AIアプリケーションは、ベクターデータストアを活用して、新しいデータと既存のデータの間のギャップを埋めています。AWSでは、様々なデータベースサービスにベクターデータストアのサポートを提供しています。これにより、開発者は新しい機能を構築する際に、データストアとベクターデータストアを簡単に統合できます。
🛠 移行とモダン化のためのツールとサービス
AWSは、移行とモダン化のプロセスを支援するツールとサービスを提供しています。DMS、Fleet Advisor、Schema Conversion Toolなどがその例です。これらのツールは、データの分析、インベントリの作成、スキーマの変換を支援し、顧客が移行プロセスを円滑に進めることができます。また、AWSは、データ戦略チームやデータドリブンイノベーションセンターなどのプログラムを通じて、顧客のビジネスケースをサポートしています。
Mindmap
Keywords
💡データベース
💡クラウド移行
💡NoSQLデータベース
💡マイクロサービス
💡スケーラビリティ
💡可用性
💡データマイグレーション
💡サーバーレスアーキテクチャ
💡キャッシュ
💡マルチテナンシー
Highlights
Barry Morris介绍了自己在AWS数据库服务中的职责,特别是非关系型数据库。
数据是AI的基础,操作数据库是数据的基础。
Amazon自身也经历了从关系型数据库到多种数据库技术的迁移。
现代数据库应用需求包括敏捷性、可扩展性和24/7可用性。
传统应用堆栈正在向微服务和解耦应用转变。
多租户架构是云服务提供商需要考虑的重要问题。
Amazon.com的数据库现代化之旅,包括从单体架构到微服务架构的转变。
客户迁移到云数据库的典型旅程,包括从EC2开始的迁移。
Venki介绍了数据库现代化的不同路径和模式。
使用数据库迁移服务(DMS)和模式转换工具简化迁移过程。
Amazon Aurora提供了与传统数据库相比更高的性能和可扩展性。
Babble Fish工具帮助减少从SQL Server迁移到PostgreSQL的复杂性。
Amazon Dynamo DB能够处理极高规模的请求,如Prime Day。
Amazon Document DB为文档模型提供了高性能的数据库解决方案。
Amazon Neptune和Timestream针对特定数据集提供了专业的数据库解决方案。
数据现代化不仅是数据库的迁移,还包括从交易型数据库到分析型数据库的转变。
AWS提供了多种工具和服务来支持数据库的现代化和迁移。
向量数据库在支持生成性AI应用方面变得越来越重要。
AWS和合作伙伴提供的工具和服务可以帮助规划迁移和现代化之旅。
AWS提供不同的计划和团队来帮助客户在数据现代化方面取得成功。
Transcripts
um so um just want to welcome you uh my
name is Barry Morris uh I I head up a
number of the database services in na
AWS uh notably non-relational databases
work very closely with a number of the
other uh database or all the other
database leaders in the
organization um so I just wanted what
we're going to be talking about today is
about databases of course data is the
foundation of AI and operational
databases are the foundation of data um
uh and most of our customers are going
through a a sort of a journey of
migration onto on on onto the cloud and
onto Cloud databases um and um and this
is quite a fun thing for us to talk
about because uh because Amazon's been
through this ourselves um amazon.com not
too many years ago was running
essentially on relational database
technology um and it got to its limits
and uh and so what we've done
subsequently if you look at the at the
at the Amazon
uh web page today the amazon.com page
behind that application is many
databases there's search technology
that's driving the search there's
there's there's key value technology
that's that's driving the the customer
feedback stuff there's graph technology
that's driving the recommendations and
so on and we'll talk a little bit more
about that in a minute so this is a
journey that we've been on in ourselves
um we want to talk through really sort
of what it means to modernize databases
on on the cloud I'm going to give you a
little bit of an overview and Van is
going to give us a bit of a a deep dive
so um excuse me the first thing is about
the these modern modern database
applications uh the requirements are are
these uh first of all agility because
your business is agile you've got a need
to move your business forward you need
your unlike old style applications that
would be built rather like a bridge
they'd be built and you move on and
build the next Bridge today applications
are live you keep working on them you
move as far as as you can um and so
agility is absolutely key to
it um the second piece scale is actually
the primary thing that drove Amazon to
move to notably adding uh document this
Dynamo DB into into amazon.com because
uh every time that we every time that we
came to a big uh retail day um we had to
do enormous planning to make sure that
the relational databases could keep up
with the likely Peak load um since we
put Dynamo DB in there we've never had
to even plan for it um because it just
scales and so that's really the the
essence of this whole kind of
modernization thing if you use the right
database for the right thing um you get
the performance you get the agility and
so forth and lastly of course um getting
to to to 247 availability even under
situations of of system failure or of
extreme load
Etc so these are the sorts of
requirements for mod modern applications
um and traditional applications don't
typ tyly do
that uh on the left you see the sort of
typical application stack that's been
around for for a couple of decades uh
and it's and it's you've got a you've
got a presentation layer and you've got
a business logic layer and you've got a
storage layer uh kind of three tier type
applications these are exactly the
applications that people are running on
premises and uh and which don't tend to
scale very well on the cloud of course
where we're going to is what's on the
right you've still got those same basic
layers they're all very different
presentation layer is probably much more
mobile oriented it's much more it's it's
it's it's voice oriented it's natural
language oriented perhaps it's targeting
special devices set top boxes in in car
and so forth um but it's the middle
layer that's that's had the big change
which is that we've moved towards
microservices uh and to so disaggregated
applications uh running on kubernetes or
whatever uh and the goal there is that
every piece of the application can not
only scale independently but that it can
have separate teams that are working on
it and they can be going through their
cicd deployment pipelines independently
and so you start getting that that kind
of application velocity it's absolutely
at the Key of it and of course at the
bottom each of those microservices will
tend to use not only their language of
choice and their Frameworks of choice
and so forth but they'll tend to use
that the database of choice and for some
of them it's going to be a mongodb style
document database and for some of them
it's going to be relational and some of
them it's going to be key value for many
of them there's going to be a cach on
top of it um and the and and what you've
got is an architecture that doesn't look
anything like that monolithic
architecture that you had uh on
premises additionally um there are if if
you're if you if if you've historically
been selling software to on premises
customers uh so you're an isv in other
words uh and you start wanting to Ser to
serve th th those that application as a
service on the cloud you've got this
issue of multi-tenancy and uh and really
this sort of multi tendency thing is
that the naive way is that you're going
to simply fire up an instance of that
application stack uh for every customer
and that's what you've got depicted here
lots of people do this it's got various
strengths and weaknesses single biggest
weakness is that you you probably going
to over provision for your smallest
customers and it's going to cost you a
lot of money um the alternative at the
other end of the spectrum um is to is to
have multi-tenancy full multi tendency
where all your customers are sharing the
same application image and database
instances and so forth um that's hard to
do as you're going through this whole
process that's something if you're going
to be doing that that's a big piece of
work that's going to be part of your
modernization in fact a lot of customers
end up with uh a lot of a lot of of our
customers will end up having um a mix of
those two things because their largest
customers have got so much more in terms
of their needs um that it's quite it
makes sense for them to have dedicated
infrastructure and then the smaller ones
to share and so there are advantages and
disadvantages to to these two sides of
it but the main message from the point
of view of modernizing applications and
certainly moving on premises to the
cloud is think about uh multi- tendency
up front it is a big part of of the
need we mentioned Amazon uh amazon.com
um this as I say is something which uh
um is is is enormous scale um the you
know this the the numbers of
transactions that we have to deal with
on a sort of on a busy day or measured
in in trillions per day um so very very
large uh data loads um and uh and we
moved from this monolith we moved to
microservices style architecture
organizationally as well of course what
we did is the typical Amazon thing of
having small teams what we call two
Pizza teams a team that can be fed by
two pizzas um driving each of these
microservices and so now you've got
something where as I mentioned earlier
we don't really have concerns about
scale where like we had before uh we
feel that we can cope with whatever the
Peaks are on on a particular day um
because of the of the modernization that
we've
done and then lastly just before I hand
over to Vani let me just sort of picture
for you what a typical Journey looks
like for a
customer um you know typically people
are still from the bottom left and
they're and they're moving from on
premises to probably to ec2 um that'll
be a kind of a lift and shift movement
they're running essentially identical
technology but now on the cloud they got
a lot of benefit from that because they
can do things like provisioning quickly
or move to a bigger machine or add
another location or whatever um and they
they get onto both of these arrows the
one of which is the arrow of innovation
velocity and the other of which of
course is cost management um and they're
onto that but and and there's some
customers that are happy to stop there
but really what we've been talking about
in terms of Amazon and and and and and
and some of the more advanced customers
they're getting to the far end they've
moved through going through moving to
manage databases where they no longer
have to do the the the sort of what we
call undifferentiated heavy lifting of
running those database services and
doing the backups and the patching and
everything else we'll take care of that
for you uh once they've done that quite
often people that came onto the cloud
running on a commercial database will
feel well I don't need to do that I can
actually run on postgress or my SQL or
something so let's me move let me move
away from that and ultimately moving to
modernization which is to adding elastic
Ash to make to because you can do that
on the cloud and it's much harder to do
on premises or adding Dynamo or adding
whatever it is um moving document DB in
because you've got content management to
do and so forth so really that's what I
wanted to do by way of introduction I'm
going to hand over to venki um and he
can take us into a deep
dive thank you thank you
Barry so perfect let's let's take a look
into the pathways for modernization so
what you have learned is customers would
like to modernize as they migrate into
the cloud right and as you all know
modernization is not a process it's a
journey and you may have different
comping reasons to migrate and most of
the common reasons we hear from the
customers are data center exit or end of
Hardware software licensing or simply
for cost
savings and the board of directors CTO
CFOs
are no longer expect uh they no longer
expect cost savings although you do
expect to save at least 25 to 50% when
they move into on uh when they move to
AWS when you compare it with on
Prem they would also want to know
transform the customer experience they
would want to digitally transform the
business and most importantly they would
want to reinvest that cost savings back
into the Innovation cycle and that is
very important and for that you cannot
modernize just the database alone right
you know you have to start with the
application no matter what so when you
start building right you know you have
to analyze your existing portfolio of
applications now a lot of customers you
know we have learned that they start
from the scratch by building you know
brand new applications mostly they use
serverless as their building blocks and
basically they are technical depth free
right you know and then they look into
the existing portfolio of applications
they will try to to either Reduce by
returing their applications that they no
longer want or they try to consolidate
by moving into multi tency they try to
safy their
applications then most of uh the
customers would like to do a lift and
shift or they migrate to AWS and the
last pattern is the modernization right
now this is where you would see
refactoring and re-platforming and this
is where we going to spend some time on
it so here is a slide which is probably
quite busy but then let's start from the
left hand side or right hand side to me
uh so that's the application you have
databases and windows right you know the
first step that most of the customers
would like to do is rehost right you
know you can rehost the application into
the cloud on ec2 for example but then
there are requirements from the customer
where they do not want to move into
Cloud probably due to some stringent
requirements uh they can go for public
CL if public cloud is not an option then
you can go for AWS Outpost you know that
is an option for customers to run AWS
infrastructure on Prim to get a full
hybrid
experience and then when you want to
completely refactor and rep platform
your applications you know you can move
into more of a cloud native architecture
as you see on the
screen so let's say you have to build an
event driven architecture or or
applications that are already being even
driven you can go for Lambda right and
then you can go for containers
especially if you have any VMS you want
to move into Cloud you can go for
container especially they're great for
monolithic applications or let's say
that you want to run and deploy long
running
applications then let's say you want to
have uh more of a controlled environment
where you do not want to spend time on
managing you could go for ECS which is
again a managed container service where
you can run deploy your cized
applications you also have foget which
is a computer engine for managing your
uh conization then you have eks where
you can again uh manage your conization
applications again that is based on open
Service open source kubernetes so that's
there so from a developer perspective
you know they would like to have more
agility right you know we do have
applications like codar where developers
can build deployment workflows on AWS
and for continuous delivery you have
code builds uh code Pipeline and so on
then moving into the data layer you know
this is where you have mve to managed
and mve to open source where either
you're going to move from let's say SQL
Server to audio SQL or let's say you
want to move away from SQL Server to
Aurora postgress so we will spend some
time into the next few slides
there so what what are the common
patterns that we normally see from the
customers right so let's say you have uh
like thousands of enrollments at your on
Prem and you want to make an inventory
the most challenging part is like you
know you need to do a manual auditing to
find out what your data sources are how
many databases you have and so on and so
forth that could be a task which could
take months right in order to make that
job easier we have a tool from the
database migration service called data
collector so with that tool you know
it's automated it's going to collect all
your inventories probably in Hearts
you're going to get a list and that list
will have all your environment with the
versions the database size and so on so
you could make an informed decision of
like you know okay which database I need
to start with and so on and so forth
then let's say you have commercial
engines like SQL server and Oracle
server and you want to make a decision
okay I want to do
migrate to an ec2 or I want to move into
RDS the first option most of the
customers see is ec2 obviously they use
native tools to migrate but let's say
they have some restrictions and they're
unable to use uh Native tools then you
have DMS and S what it stands for is
database migration service and schema
conversion tool so the schema conversion
tool is probably the first step where
you know it's going to create an
assessment report and if you're unsure
let's say I don't know where to migrate
to from s you know should I pick my SQL
or should I pick postris then schema
conversion tool would give you an
assessment report and that tool is going
to tell you okay this is the best Target
because the level of effort for you to
migrate is probably the least so with
that you're going to make a decision and
not only that it it's going to also help
you do schema conversion it's going to
do code objects conversion and the most
important part is the data migration
which is taken care of by TMS and for
minimal downtime or near zero downtime
it also helps with data replication as a
part of DMS what we call it as
CDC the first pattern is pretty obvious
that is M to manage so let's say you
have an nonpr SQL you then to uh RDS SQL
servers which is pretty straightforward
you can use DMS in or native tools of
your choice to migrate which is
great and uh just before going there
like you might wonder you know what is
the benefit of moving to a manage
service you know like Barry said manage
services are great if you don't want to
spend your time on undifferentiated
heavy lift like you know provisioning
patching upgrades security and so on so
R is as a managed service we give you
that experience so that you can focus on
Innovation the next pattern which we
commonly see from the customer is breakf
free that is you know I want to move
away from SQL to something else and this
is where sat is going to help you you
know it's going to create your this
assessment report and you're going to
move from there so in this case you know
I picked Amazon Aurora so for those who
is not aware of arur uh so this is
compatible with both my SQL and postris
which gives you five times more
performance than uh vanilla my SQL and
three times more performance than
postris and the best part is you know
the storage is distributed um you know
it's automatically uh scalable and so on
and so forth so I have two more uh
elements there which I will cover after
this so that's the babble fish again for
those who is not aware of Babble fish
you know moving away from SQL Server to
let's say aor postris is a task on its
own right you know it's timeconsuming is
resource
intensive but with Babble fish we have
reduced the downtime or reduced the
effort so what it means is uh let's say
you can keep your applications that are
written for the SQL Server post can able
to understand so basically it can able
to understand both tsql and psql so for
the new codes that you're developing you
can continue writing it on psql and for
the existing codes you can take your
time to convert and until then you can
continue to use your tsql codes so it's
basically a machine that can understand
both
languages and the last most popular
option is the purpose B database because
what we understood after working several
with the customers is that there is not
one database that suits for all your use
cases right you that isn't a database so
although Amazon Dynam DB is what I've
used in the screen uh there are multiple
databases which I will be talking about
in the next few screens and before I go
there I want to talk about Aurora
Limitless for a bit so you know
relational databases can one scale to an
extent right whereas aora Limitless
gives you an option to scale almost
infinitely right you know you with the
writer in node for example you can
almost able to achieve million
transactions per second you know that is
unbelievable for a relational database
yes you can able to achieve that with
purpose build databases but from a
relational perspective normally you are
bound to hit some of the uh capacity or
the limitations which we have taken a
feedback and worked on a new tool that
is Aurora
Limitless So within this modernization
uh pattern right you know what we' have
seen is adding this caching yes you can
work with the databases uh normally what
happens is the data is going to be
fetched from the disk which is an
expensive process we going to put it
into the memory and then you're going to
read it from the memory which is all
right but then as you keep doing you
know that is going to add performance
penalty to your applications but let's
say your developers need microsc latency
as a basic requirement then the
relational databases can't do
justification right and this is where
AWS has given you an option that's
elastic cache in memory DB now Amazon
elastic cache is R is and M cache uh
compatible so persistence is an option
so you can enable or disable and it
gives you microsc uh response times and
on the other hand uh Amazon memory DB is
fully compatible with redis and it's a
persistent inmemory data store right so
you get uh you know microsc latency for
reads and millisecond latency for rights
and keep in mind you can use this
database as a persistent inmemory data
store especially where the application
sensitive application is Sensi to
latency this is great and as you see you
know you can achieve scalability and
Agility with the help of a purpose build
database like elastic cache and memory
TV then the second subpattern again for
non- relational uh I mean for the
purpose buil is a non-relational query
patterns yes you know you can scale to
an ex with the relational databases but
what if you have uh a requirement where
you have to have an high performance
application that you know that has to
scale infinitely right such as
amazon.com you know when when you have
this
amazon.com uh sales like Prime day for
example uh just to give you a stats
Amazon Dynamo DB was able to serve 45
trillion requests a day so that is 45
million request per second sorry 7
trillion million requests uh a day and
then 45 million requests per second so
that's a lot imagine you know we don't
expect customers to be scaling of that
size but what I'm trying to say is you
know you have a service that can able to
scale that much uh so Dynamo DB is built
for key value database like I said it
can able to perform at any scale and uh
so if you have an use case where your
developers needed uh you needed you can
go for 10B
and the second popular engine is the
document TB in this diagram uh so again
this is a purpose buil database for the
data document model so let's say you as
a company that's working on a lot of
documents be it nested documents that
you'd want to query the document or
index the document uh then doing it on a
relational uh database is probably timec
consuming right you know you have to
normalize that you have to fit it on a
relational table and that's not going to
scale well and this is why we have come
up with uh Amazon document DB which is
fully compatible with mongodb so which
means any apis drivers application codes
that you're using already with mongodb
you can able to use it with Amazon
document DB and some of the benefits
again same as what we seen before
scalability agility and
performance and uh the last uh is the
specialized data sets what we call so
these are like unique cases right so
let's say you as a company that uh that
is working with highly connected data
sets uh let's say you are a company
which has millions of followers and you
want to establish a relationship between
the followers and let's say an athlete
for a sports store then uh you know
social or sorry the graph database such
as Amazon Neptune is a great fit for
your use case and for uh time stream
imagine you I have a data set uh based
on time that's constantly changing and
you have like uh GBS of data that's
getting ingested per minute uh obviously
in relational database can't do
justification for that especially when
you want to create index on a relational
database for that size the indexes are
going to be bit clunky right so you
don't uh you know it's not going to
scale well so this is where Amazon time
stream is going to come in really handy
and this is a database where you know
you can ingest tens of GBS of data
permanent and you can able to create
that in fraction of
second so that's the whole uh uh in a
purpose built diagram with different use
cases and now let's see what customers
do let's say after migrating right so
you know you have your applications now
as a customer you have made your case
you have migrated you have probably
modernized some of your
databases uh the next step
is what to do with this data right now
data is a foundation right so you would
want to move some of the data from the
transactional engine to the analytical
engine and probably beyond for AML
Innovation and for that what you seen is
you know customers normally uh have
these developers they would want to use
new services but most of the times when
they want to use this new service or a
plugin they would have to
rearchitecturing
a service anywhere you can build a
pipeline and you can move the data
around so for streaming we have Amazon
kindnesses and uh
msk and for batch Eng you have uh
services like TMS you have Amazon EMR
and you can use open source Big Data
Frameworks like Hive spark Presto and
you also have glue where you want to do
an ETL kind of you know extract
transform and
load and then you move your data build
your analytics or your data L of that
sort and then from there you know you
can do a visualization using Amazon
quick site which is a business
intelligence tool or you can like you
know stream it back to other backend
applications or you could basically use
this data to build train and model your
uh machine learning llms right and
that's where Amazon sagemaker comes in
uh handy and you do also have Bedrock
know we have a separate track for that
so I'm not going to dive deep into that
and one other area which I would want to
highlight is the ZL uh here so ZL is
another platform that's going to help
you move your data with a click of a
button from your analytical database
sorry from your relational database to
the analytical database so from Amazon
Aurora to Red shift uh you have the zero
zero ETL integration that works uh
smoothly uh our next session is on zero
ETL so I'm not going to jump too much
into it so you can wait and listen to
the next session to uh get more about it
so these are some of the zero ETL
features that's available as of
today and again now it's easy you don't
your developers don't need to spend time
building this whole Pipeline and
something breaks you know you don't want
to spend time on investigating and
fixing it uh we made your life easier
and what's even interesting is quite
recently we have made uh a new feature
where you can take control over your
Source database on like which tables
that you want to migrate it from olp to
oap so that's more on the next
session and if you want to go seress
obviously you have a lot of options on
the application and here is a snapshot
of the applications from the database
side you have all this that's
highlighted in white uh or the database
layer of course so you can use uh Sous
offerings
here so what's next yes you have
migrated you have modernized you have
innovated like you know uh what's the
next popular pattern that we have seen
right you know with this whole whole
generative era we have seen Vector
databases so before I come in here I
want to talk about this slide I'll go to
that one later so probably have seen
this uh slide before but I would want to
focus just on the vector data store
right you know so Vector data store is
an emerging thing uh especially that
supports this generative AI
applications so again we have a separate
topic running on this Vector but just to
give you an overview let's say you as a
customer or present with a prompt you
enter uh a question and you're looking
for an answer so what it's going to do
is it's going to look into the llm and
it's going to fetch and it's going to
provide a context as an output but what
if the data that you have trained uh was
trained on let's say 3 months ago right
know and there's a lot of data that's
probably has been populated right after
that and if you asked any question of up
toate data probably you're not going to
get any up toate information so how are
you going to tackle it and what we call
it as h ination in in a world so in
order to you know tackle this
hallucination problem you have this rag
based solution or retrival augumented
generation and to support that you have
Vector data store so basically you're
going to create embeddings and you're
going to store into a vector database
Ting is nothing but a format that
machine can understand to train uh so
with this what it's going to happen is
uh with all the new data that you're
getting from the right hand side we're
going to create embeddings we're going
to put it into the vector data store and
for the questions that's being asked
from the user we're going to use the uh
context and we going to do a similarity
search we're going to get the context
and with the given input we're going to
add this additional context to the
ground Alum so that it gives you an
accurate information and this is where
Vector databases are popular so I'm
going to go back to the previous slide
and then I'm going to jump onto that so
you might wonder like you know why
having this vectors and data store uh
are important like why do you want to
keep it together yes you do have a
separate uh Vector data store but then
as a developer you don't want to build
another pipeline to move your data from
your data store to another Vector data
store right you want to keep everything
in sync so that you don't you avoid this
data moment your developers probably
will have to learn a new language to or
create a new apas and so on and so forth
so you're going to avoid all of this if
you are keeping your vector and your
databases together so to make that
easier I'm going to move now on to the
next one sorry for
that so we have all of these services
that you on the screen that supports
Vector database now many of you have
already used or some of you have already
used the service so we have Vector data
store support in all of it right so you
can continue using your database as you
do and then if you have a new
functionality of you are building a new
generated U application then you can
continue using the vector data store uh
in this
service so
now what are the tools that's available
you know we have spoken about everything
now what are the tools that's available
so you have DMS Fleet adviser which you
already seen that helps you basically
doing an inventory analysis and you can
create a business use case basically to
perform this migration and you have
schema conversion tool that helps you
assessment report schema conversion code
conversion and so on and the database
migration service itself which would
help you to remove the data and helps
you replicate then you know we have
folks like us essay Partners in the room
and outside and uh again internally we
have dmas then you have a lot of other
programs which you'll be hearing
throughout the session here in other uh
areas so I'm not going to spend much
time on
it uh and we spoke about schema
conversion tool so I still add a slide
here so why is it used like I said you
know create this assessment convert
schema code objects and not only for
relational database as you see on the
far right it also helps you extract data
warehouse uh data to Amazon red
shift so what SS and Target to DMS
support as of today so here is a
snapshot of that uh and this list is
growing so you can take a picture and
when we show this probably you'll have
new sources next
time so when we speak with the customers
right now we start with the data
flywheel I mean you may have a different
data flywheel but then we work with the
customers we create uh you know their
own custom data flywheel and if you Noti
is you know it's made up of few key
initiatives and one key initiative
doesn't work on its own it has to be a
combination for the business to grow in
long term uh so the businesses that
actually Focus is on each part of this
flywheel has a long momentum right you
know this is where AWS will work with
you Partners will work with you in
creating this flywheel and we help you
succeed so what are the takeaways you
know like I initially said modernization
is not a process but rather a journey
and you can use the pathways and decide
like you know where do I start and where
I am and then from there you can move
your next step because you know it's
it's it's a journey like we said and
with tools that we have shown from AWS
as well as from Partners you can start
planning your migration and
modernization
journey so we have a lot of programs uh
again you would be seeing this in almost
all the sessions but just to give you a
high level overview if you need P level
executive guidance we have data data
strategy team so they are basically a
former cxos from large Enterprises we
will team you up with those uh members
and we can help you create a business
use case and we have D2 or data driven
everything which is again a program
where we'll work with your C Level uh
and we'll create a business use case or
ideation day and we will help you uh
achieve that uh for a long-term
progress then you have reimagine data
and you have AWS generative AI
Innovation Center where we will again
team you up with uh you know folks with
the skill set on that particular domain
and we will help you achieve what you
want to build in the uh next in ation
cycle on the generative
Ai and with that I'm I'm going to
conclude the presentation thank you very
much for listening to our special thing
I appreciate that
Voir Plus de Vidéos Connexes
Multicloud Development with Microsoft Azure and Oracle Database Cloud | Oracle DatabaseWorld AI Edit
PerplexityのPro検索の凄さがやっとわかったので共有
盲点!Difyのローカル版でなく、クラウド版を使ったら、RAGチャットが普通に公開できちゃったし、回答も鬼早だった
【B6】Copilot for Microsoft 365 で実現する未来の働き方とその準備のポイント
【ヒット商品の作り方】トップクリエイターが教える「言葉の経営術」。「2単語コンセプト」「常識の真逆」「矛盾の組み合わせ」で価値を生み出せ(細田高広:コンセプトライティング)【NewSchool】
[番外編 #02] ペルソナ、ちゃんと使えてる?番外解説編(前編)
5.0 / 5 (0 votes)