Girish makes a good point. To get started with AWS, you simply click a link and it takes your AMZN credentials. So there could be a lot of tire kickers. What AMZN isn’t showing is number of instances, though the have stated publicly that the amount of traffic on their web services exceeds the traffic they are putting into their own system. That’s impressive, though if I were being cynical, I would think that was for a low traffic month (i.e. Feb), and not for an entire year (which has the holiday rush of Nov and Dec.).
Microsoft is a developer platform company with millions of developers, and may billions of dollars in revenue created by those developers building on our platform. We’re greatful to have them, and bringing the right ones over to Azure is a top prirority.
The 500K # for AWS is total developer accounts created...don’t think that everyone is developing an app. AWS has a head start....but if Azure is the right platform - many from the existing(several million?) MSFT developer community will be on-board. Girish
Thanks for the comment. I was more speaking to AMZN as a company and as a developer ecosystem provider.
As to your question about middle tier apps, if you think about what the cloud is supposed to be (or rather, could be in our wildest dreams), the middle layer is actually abstracted away and handled by the system. Re: Oracle, I would love for them to port their app to the Azure ecosystem. If you happen to have a line to Larry, let me know. :) The more database options in the cloud, the better, for all parties.
We are going to talk about a new technology. Many will fundamentally re-think what computing means for their organization. A technology that is, for the first time, democratizing IT by making it plentiful, standardized, and always available.
Cloud Providers Public 030909 V2 - Presentation Transcript
What’s Going on in the Cloud? Brandon Watson Director, Azure Services Platform [email_address] www.manyniches.com twitter.com/brandonwatson
Today’s Topics
Who am I?
Re-tread at Microsoft
Ex-private equity guy
Entrepreneur on loan to MSFT after my last exit – “wifestyle” reasons
My last company ( IMSafer ) was an exit for cash
Why cloud?
Continuing trend of simplifying app creation & service delivery
Further reducing friction in creating new opportunities for business
Concentration of IT dollars
Who are we talking about today?
Amazon Web Services, Google App Engine, Microsoft Azure
Not VMWare, Salesforce.com or multitude of startups
Future directions?
Investment whitespace and needs
First, Some Definitions There’s a lot of aaS’s out there
Infrastructure as a Service (IaaS)
The ability to manage and launch virtual machines on hardware that you do not own
Platform as a Service (PaaS)
A more abstracted view of the development platform, where all facilities required to support app life-cycle are available via network
Software as a Service (SaaS)
The delivery of an experience to a customer, on demand, generally using the web as a delivery mechanism
Software Plus Services (S+S)
Using the best of software and services to deliver the right experience to customers at the right time on the end point of their choosing
Brandon as a Service (BaaS)
Rare Seattle fish, known for bloviating and biting
Cloud Computing Platforms
My themes for today
Functionality
How much can you do with what you have?
Familiarity
Do you instinctively know how to use it?
Scalability
How easy is it to handle more inputs and workloads?
The Trouble With Big
Large scale deployments generally get one of these two reactions
Cloud to the rescue!
Amazon Web Services Service Types Infrastructure | Payments
Elastic Compute
On-demand, resizable compute
Elasticity is not automated
Supports Linux & Windows server
Very easy to get started, and use
Simple DB
Core DB functions (index and query)
Schema-less DB, much like BigTable
Limited primitive types
Optimized for CRUD operations
S3 Storage
Simple web services for storing arbitrary amounts and types of data
Data store is unlimited (objects to 5GB)
It’s about availability over consistency
Dev Pay
Billing as a service
Ensures authorized app access
Flexible Payment Service
Payment system built on top of AMZN’s infrastructure
Customers can use AMZN credentials for payment
Simple Queue
Service for storing and passing messages between computers
Queue messages are durable
CloudFront
Web service for global content delivery
Targeting Akami with lower cost
Datacenters Are Spendy
Best to share that risk if possible
Source: AMZN 10Qs & 10Ks
5 Theoretical Whys of AMZN
Why is our CapEx so high?
Because we have to buy a lot of servers.
Why are we buying a lot of servers?
Because we need to support the traffic needs of our business and our partners’.
Why aren’t we able to support them with fewer servers?
Because traffic patterns and server utilizations require more servers.
Why do traffic and utilization patterns have an impact on number of servers?
Because the traffic spikes in Nov and Dec create a need for peak load.
Why does the traffic have massive spikes in Nov and Dec?
Because AMZN and all of it’s partners are in the same business – retail.
Resolution – attract customers onto your technology platform who have different patterns, and no longer are major traffic component in own system
Amazon Web Services
Hosting++ is more expensive than you think
My last company as an example
Hosting with AMZN would have cost $4,500 / month
3 front end servers, 1 dev front end server
2 database servers, 1 dev database server
2TB of storage
Bandwidth
10U of hosting cost us $750/month
Hardware is cheap and getting “free-er”
Net out
Scalable – Scales nodes, not algorithms
Functional – Good UI with lots of services available for building apps and services
Familiar – Use the server apps and OS that you already know
No real developer DNA, and no support for partners
Google App Engine Service Types Infrastructure
Compute
Highly scalable and built on top of GOOG infra
Elasticity is automated
No concept of an OS or a virtual machine
GOOG limits the actual scalability
BigTable Storage
Entity relationship storage
Built on top of GOOG’s database for their site
No tools for data management
5 Theoretical Whys of GOOG
Why are we making money?
Because we sell advertisements against pageviews.
Why aren’t we selling more advertisements?
Because there is a lack of usage of online services globally.
Why is the growth of online services stalling?
Because there is a lack of content, not connectivity.
Why is there a lack of content?
Because there is no cheap and easy way to build and host dynamic, big scale applications for the web.
Why is there no easy way to build dynamic, big scale applications?
Because hosting can be expensive, and setting up servers for large scale can be extremely difficult.
Resolution – release a low cost, highly scalable hosted application programming framework with a very specific target application type in mind – to create content against which to sell ads.
Google App Engine
My First Web App
My last company as an example
Could not have handled the load – too many transactions per second
Could not have handled any of the background processing
Complete re-write of data-access layer
Complete re-write to Python
Net out
Scalable – Rate limited by GOOG with little visibility into what resources are in use
Functional – Very limited by features made available
Familiar – Language, framework and database all new concepts
No SLAs, but marketplace is coming for GAPE plug-ins
Azure Services Platform Build Apps & Services Using What You Know
Azure Services Platform Service Types Infrastructure | Connectivity | Security
Elastic Compute
On-demand, resizable stateless application and service hosting environment
Elasticity is automated and managed by fabric
Supports multiple programming models
Simple queue built in
Very easy to get started, and use
SQL Data Services
Core DB functions (index and query)
Relational in the cloud
Built on top of Microsoft SQL
Table/Blob Storage
Simple storage for the compute layer
Data store limited and not meant for heavy DB
Tables and blobs for storing many data types
Service Bus
Easy connectivity thru firewalls
Passing messages between apps and services
Queue messages are durable
Multiple connectivity options
Access Control
Standards based
Claims based authentication for connecting users to apps and services
Live Services
Easy synchronization of files, folders and applications
Access to MSFT social graph
Great developer services and monetization engine
Azure Services Platform
Breadth and depth to deliver solutions
My last company as an example
Ruby & Rails support is “coming”
Might still need a RDBMS local to Ruby code (depends on SQL Data Services issue)
Message queuing would have helped immensely
SQL Data Services would have solved our database sharding challenges
Much easier to connect to on-premise deployments (our B2B solution)
Net out
Scalable - Easier to build scalable apps and DB, and costs look like they will be lower due to utilization optimization
Functional - Many of the core on-premise developer features which customers depend on are available
Familiar – Standards based and can leverage existing developer skills
Deep support for ISVs and partners to build a business
Ideal Cloud Scenario
Deploying on the cloud should feel like this…
Whitespace
What’s interesting, and what’s missing
Essential Services
Very hard for a startup to compete in building out a new operating system or development platform layer
Building Block Services
Developers will need plug and play pieces for services they are building – PHP is a great proxy for how this could shake out
Billing, data management, business intelligence, application bridges
Finished Services
Plenty of SaaS applications are being created – estimated $20B in revenues already
Lots of opportunity for bridging on and off-premise enterprise applications
4 comments
Comments 1 - 4 of 4 previous next Post a comment