17:06:33 <jmeeuwen> #startmeeting Cloud Computing
17:06:33 <zodbot> Meeting started Sat Dec  5 17:06:33 2009 UTC.  The chair is jmeeuwen. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:06:33 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
17:08:20 <jmeeuwen> most computing details are preconfigured and managed by the cloud
17:08:26 <jmeeuwen> hardware setup, management
17:08:34 <jmeeuwen> common software services and images,
17:08:37 <jmeeuwen> remote storage services
17:08:40 <jmeeuwen> #chair gregdek
17:08:40 <zodbot> Current chairs: gregdek jmeeuwen
17:08:58 <gregdek> Max Spevack: diff between cloud and grid?
17:09:36 <loupgaroubl0nd> jmeeuwen, cheers
17:09:40 <gregdek> Morsi: Grid is centralized and uses a thin cloud, cloud is generic and more flexible -- think of cloud as successor to grid
17:10:59 <jmeeuwen> #halp
17:11:29 <gregdek> Cloud features:
17:11:32 <gregdek> * Simple to get an OS up
17:11:37 <gregdek> * Simplified mgmt interfaces
17:11:43 <gregdek> * Standard install scenarios
17:11:51 <gregdek> * Wehosting ++
17:11:58 <gregdek> Use cases:
17:12:02 <gregdek> * Web Hosting ++
17:12:13 <gregdek> * FOSS communities can scale resources up or down as needed
17:12:24 <gregdek> * Lone developer has a way to get resources easily
17:12:35 <gregdek> * Can mothball a project easily and resume later
17:12:41 <gregdek> * Research / Academic
17:12:44 <gregdek> * Geographic diversity
17:12:56 <gregdek> * Joint ventures
17:12:58 <gregdek> * Test bed scenarios
17:13:09 <gregdek> * External auditability
17:13:44 <gregdek> Good cloud computing paper:
17:14:05 <jmeeuwen> #link http://www.cs.ucsb.edu/~lyouseff/CCOntology/CloudOntology.pdf (very nice paper)
17:14:09 <gregdek> Extends from many existing concepts:
17:14:15 <gregdek> * Distributed/Grid computing
17:14:36 <gregdek> * Service Oriented Architecture.  For cloud, *everything* is a service.
17:14:42 <gregdek> * Virtuzalization.
17:15:36 <spevack> Mohamed Morsi is talking about Cloud computing.
17:15:41 <spevack> Right now it's pretty much just generic stuff
17:15:47 <spevack> He has recommended the following link:
17:15:49 <spevack> http://www.cs.ucsb.edu/~lyouseff/CCOntology/CloudOntology.pdf
17:15:52 <gregdek> (We got it, Max)
17:16:04 * spevack shuts up
17:16:07 <jmeeuwen> nice: HaaS, CaaS, DaaS, IaaS, PaaS and *then* finally SaaS ;-)
17:16:39 <gregdek> Definition: "A new computing paradign that allows users to temporarily utilize computing infrastructure over the network, supplied as a service by the cloud provider at possibly one or more levels of abstraction."
17:17:09 <gregdek> Cloud providers:
17:17:13 <gregdek> * Amazon
17:17:16 <gregdek> * Google App Engine
17:17:19 <gregdek> * Rightscale
17:17:21 <gregdek> * Rackspace
17:17:27 <gregdek> * RHEV (private cloud)
17:17:33 <gregdek> * oVirt (private cloud)
17:17:35 <gregdek> * Many more
17:18:12 <gregdek> MANY APIs, ONE PROBLEM:
17:18:20 <gregdek> * Each cloud provider has their own API!
17:18:53 <gregdek> * Therefore, cross-cloud mgmt and migration are problematic.
17:19:02 <gregdek> * Which means people are nervous about adopting.
17:19:19 <gregdek> DELTACLOUD!
17:19:21 <jmeeuwen> #link http://deltacloud.org
17:19:35 <gregdek> * open source API and framework that abstracts difference between clouds
17:20:00 <gregdek> * abstract interface to manage and access any number of cloud services w/ability to add new services with simple driver interface
17:20:10 <gregdek> * web ui (deltacloud portal) providing
17:20:16 <gregdek> + web access to api
17:20:19 <gregdek> + instance grouping
17:20:28 <gregdek> + central mgmt
17:20:32 <gregdek> + billing
17:20:36 <gregdek> + quotas
17:20:38 <gregdek> + monitoring
17:20:41 <gregdek> + and more
17:21:19 <gregdek> (which will be available in the slide deck)
17:21:50 <gregdek> * Framework is REST based
17:22:00 <gregdek> * Out of box cross language support
17:23:03 <spevack> #link http://en.wikipedia.org/wiki/REST
17:23:34 <jmeeuwen> slide shows a little ruby script
17:23:38 <jmeeuwen> !RUBY!
17:26:29 <gregdek> Another script
17:26:36 <gregdek> And then a screenshot of the web UI for deltacloud
17:26:47 <gregdek> And another screenshot
17:27:01 <SmootherFrOgZ> it's ruby rails app
17:27:25 <gregdek> * mock driver exists
17:27:36 <gregdek> * source, rpms all on deltacloud
17:27:58 <gregdek> * working on getting deltacloud into fedora, not there yet
17:29:11 <gregdek> * early in dev cycle, encouraged to build from source for right now
17:29:29 <gregdek> * git clone framwork, drivers, portal, ruby client
17:30:17 <gregdek> * if you set up web portal, set up database (pgsql, sqlite, etc.)
17:31:47 <gregdek> * specific commands for starting things up
17:32:11 <gregdek> DELTACLOUD ROADMAP
17:32:16 <gregdek> * deploy same image to many clouds
17:32:21 <gregdek> * cross-clou dmigrations
17:32:24 <gregdek> * load balancing
17:32:28 <gregdek> * instance stats collection
17:32:35 <gregdek> * monitoring and alert support
17:32:45 <gregdek> * expand cloud drivers
17:32:51 <gregdek> * add quotas and billing layer
17:34:09 <gregdek> Q by Matt Domsch: how is deltacloud different than eucalyptus?
17:34:48 <gregdek> A: eucalytpus is private cloud + amazon only.  deltacloud will be an api to manage multiple types of public and private clouds.
17:37:19 <gregdek> END OF DELTACLOUD TALK.
17:37:34 <gregdek> Speaker: Justin Forbes.
17:37:49 <gregdek> Where fedora is re: ec2 and where we need to be.
17:37:53 <gregdek> Justin at whiteboard.
17:37:59 <gregdek> Fedora now:
17:38:06 <gregdek> * Run an F8 image!
17:38:18 <gregdek> * Or run an Ubuntu image!
17:38:29 <gregdek> * Amazon is embarrassed and wants to support Fedora now.
17:38:43 <gregdek> Starting with F13 and every release thereafter:
17:39:01 <gregdek> * Create regular F(n) CDs and spins;
17:39:18 <gregdek> * Also create F(n) Amazon Machine Images (AMI)
17:39:25 <gregdek> Three parts to defining an AMI:
17:39:51 <gregdek> 1. The AMI filesystem (basically raw disk image);
17:40:09 <gregdek> 2. The AKI (kernel image, can ONLY used published kernel images, more on this later);
17:40:15 <gregdek> 3. The ARI (ram disk image)
17:40:39 <gregdek> All defined in an XML file.
17:41:00 <gregdek> Matt Domsch: could we do it sooner?
17:42:11 <gregdek> David Huff: we can do it manually sooner, but want to get it automagic.
17:42:37 <gregdek> Greg DeK: if we have a hackfest, can we have more people figure out how to help?
17:42:51 <gregdek> Jesse Keating: We've already talked about how this will fit into the release process, it's all good.
17:43:36 <gregdek> Matt Domsch: can we get an update server for Fedora inside of Amazon?
17:43:41 <gregdek> Justin: yes.
17:48:36 <gregdek> #action gregdek: figure out how to get a developer account asap
17:49:44 <mdomsch> need to ensure corresponding source is available for all packages in the image
17:50:11 <gregdek> #action jesse keating Ensure corresponding source is available for all packages in each AMI
17:50:39 <gregdek> Seth Vidal: once image is deployed, can we have post-install steps in EC2?
17:50:53 <gregdek> Justin: not yet (maybe something deltacloud could do?)
17:51:17 <gregdek> Justin: amitools has a plug-in system, so there are options
17:52:21 <gregdek> Jeroen: can we just create AMI/AKI/ARI based on kickstarts now?
17:53:35 <gregdek> Justin: we need an "official disk image" to base off of
17:54:17 <gregdek> Gregdek: it's easy to build custom AMIs, many people do, so long as the right AKI is present.  Therefore, it will be easy, once there's an "official" Fedora kernel in EC2, to build all kinds of tools for spinning up custom AMIs.
17:56:12 <mdomsch> http://alestic.com/  lists the Ubuntu and Debian AMIs available in EC2 today
18:02:57 <gregdek> #endmeeting