sdforum: cloud services SIG 2 (cloud wizard)

– meta

— opensource cross-cloud scripting

— written in python

— floats over amazon, mosso, gogrid, etc.

– why python?

— java’s written by committee

— 3000+ packages already written

– design

— cloud object

— authentication

— lists services available

— config files

— service obj

— list to see the types

— device obj

— reboots and deletes

— runs file services

— shell access

– features

— storage: s3, nirvanix

— DB: SDB, Bigtable, etc

— will also run dreamhost pseudo server

– contribute

— bsd license, which won’t go away like a gpl license

– QA

— advantages of python? shell-like

— choices for management tools?

— rightscale, cohesiveft, ylastic, scalar, cavoo, coolparty, elastra

sdforum: cloud services SIG 2 (cloudfoundry)

– meta

— chris richardson

— automated, outsourced, data center management for java apps on amazon ec2

— sdforums does cloud camps?

— wrote “pojos in action”

– why java?

— 22% market

— grails?

— spring, hibernate, terracotta, scala

— java’s community is growing

– cloud

— 18 months experience

— learning curve

— ephemeral storage

— elastic ips

— cloudtools?

— provide amis preconfiged w/ java stack: apache/tomcat/mysql

— cloudfoundry

— a hosted service

— targeted at jvm community

— monitoring ans automated management

— support

— features

— you provide war files + sql scripts

— written in jvm language

— portable

— run your apps on your amazon instance

— you can use your favorite framework

— demo

— define a name

— upload a .war file

— .war file is stored in s3

— example

— grails app

— short-term

— fluctuating load: sat/sun 4 servers, mon-fri 1 server

— a perfect candidate for amazon ec2

— app deployed by the developers

— conclusion

— contribute to cloudtools

— tiers

— db

— terracotta caching

— ap server

— ?

— QA

— what if my app needs to run solo? not yet

— ejb/j2e apps?

—- you shouldn’t be using it.

—- not yet

—- ibm will build AMIs for pop software

sdforum: cloud services SIG 2 (rightscale)

pre-talk

– next month: hypervisor showcase

— ms hyper-v

— vmware

— citrix xensource

– upcoming

— 3/3 java sig: running java apps on ec2

— 6/3 security sig/cloud services

— 40% off etech w/ code: et09ffd

— oscon in san jose

— interop

— salesforce talk

— cloudcamp?

— new version of app engine released last night

 

right scale

– michael cramdell, ceo of rightscale

— cloud deployment platform

— 10k users

— 300k servers

— features

—- automated management: reduced admin overhead

—- cloud neutrality

—- transparency

— sits btwn amazon, gogrid, rackspace, etc and cloud apps

— customers: animoto, sling, woot

– history

— pre-public w/ amazon cto 

— taught grad class w/ ror apps launched on ec2, but students would leave servers up.  so rightscale was built to automate instance spin up/down

— motivation

— forking a server

— pay as you go. foundations google and yahoo have built(!)

— just launch a server

— cost analysis

— 1 server for 1 yr on ec2 = $876

— 1 programmer for 1 day = $400- $1000

— throw servers not humans at problems

— classic dilemmas

— over building hardware = wasted money

— peak vs. avg utilization

— new mindset

— plan for failure

— focus on automation

— servers are disposable

— great for business

— 45 day float on bill

— no need to buy hardware

— no techcrunch

— ideal for spike-y usage

— barriers to customers

— fear of lock-in

— security

— loss of control

— overhead of moving

— customer needs

— fast onboarding

— expertise

— portability

— layers

— saas: most lock-in

— paas

— rightscale

 

– overview of deployment

— 4 x app servers

— launched dynamically

— httperf used to simulate load

— rule based scaling

— cloud computing is built on the shoulders of virtualization

— amazon uses syslog to create persistent logs from virtual servers

— cost = time to change x live images x time to retrieve, launch, change, x stored images

— server templates are base images w/ dynamic personalities, app server, web server, grid comput, etc.

— centos base images 

— on launch, download a list of programs that define personality

— clusters are based on templates, not hardware

— backend processing

— animoto has a web front end that farms off video rendering to grid backend

— failover

— for DBs,  promote slave and launch new slave

— amazon’s ebs volumes?

— mmulti-zone deployment

— all pop stacks supported

— animoto scaled from 50 to 3500 servers in 3.5 days

— clouds surpase traditional hosting

— multi server sizwes

– reasign ips

— multi failure zones

— multi geo

— dosk volumes

— layered arch

— cloud neutrality

— automation engine

— temopaltes

— apis

— web ui

— agile deployment

— partner oriented company

— ecosystem:

— rightscale under – customers, software vendors, cloud providers

— email michael@rightscale.com for a free tshirt

– qa

— do we need an account w/ amazon?  yes. 

— whats are the common features in the base image? 

— ec2 -> launch a server

— bigtable, sdb -> proprietary = lock-in

— scaling relational dbs?

— sharding dbs

— not yet fully solved

— most customers use traditional relational DBs for legacy reasons

— new datastores are being used for more specific apps

— rightscale like azure fabric controller? not really clear

— security compliance for enterprise customers? 

— clouds aren’t the best yet

— saas70?

— legacy apps?