Me and some colleagues attended cloud camp un-conference in Vienna. Here are my notes. So far I haven’t seen any slides posted. I add them as soon as they come up.
cloud computing 101
-
Alexander Wöhrer (FH St. Pölten) and Mario Meir-Huber (Codeforce)
cloud: mixture of lot of services and vendors
we have seen already lot of problems of services – those can be prevented
technological revolution? evolution from grid computing? evolution from virtualisation, servis oriented architecture
+ –
bussines revolution- cost reduction
- long term IT investments
- flexibility & scalability
- pay per use
- => standrardization of IT
+ –
requirements which are changing- sync/async
- local/distributed
- own/use
- compute/data intensiv
- alone/collective
- + –
horizontal/scalability- + –
scalability = scale up- vertical
- + –
flexibility = scale out- horizontal
- multiple servers
- also to be able to pay that
- + –
- + –
PaaS landscape- + –
layers- abstraction
- modeling
- productivity
- + –
sourceforce.com, orange scape- lot of abstraction
- metadata level
- + –
appengine- heroku
- framework
- + –
azure/amazon- instance level
- have hands dirty and know all about implementation
- + –
- + –
design best practices- + –
design for failure – system will always fail-
leverage availability zones
+ –
test it – see if you can loose your most important server and see if you can operate in 30minutes (you have processes in place for that)- Chef and puppet to automate deployment
- thow out main engineer from the window
-
- + –
- + –
security- + –
shared responsibility- + –
provider- certifications
- + –
app level- custommer responsible
- + –
tools from provider which you can reverage- firewall
- + –
- + –
- + –
E= m2 c- efficiency = migrating to cloud
- next cloudcamp 10.10 @ Uni Wien
- + –
Sam Johnston – Director equinix (100 datacenters over the words)- google/facebook one of their customers
- + –
moving from IT as product to IT as service- utility computing
- + –
distruptive innovation- mainfraim
- PC
- cloud
- evolving very fast
- + –
airplain engine annalogy- well tuned proclaim engine until boing 707
- + –
market for 5 computers (ibm 1943)- today called Amazon, Microsoft, VmWare, Google and SalesForce
- + –
why cloud?- network analogy
- + –
benefits- cost of 100 servers vs 1000 servers = per server 1/5 TCO
- homogenous (vs heterogenous enterprise datacenter)
- + –
cloud classification- public
- + –
private- still might be multitanent
- + –
community- specialized (goverment)
- + –
hybrid- mixed
- + –
reliablity- + –
unreiable soft on top of reliable hw- example: rational db
- cloud: reliable software on unreliable hw
- + –
unrealiable soft on unrealiable hw- thats why azure/amazon outages
- + –
reliable soft on realible hw- uneconomical
- cassandra on vmware
- + –
- + –
redundancy- + –
legacy: 2n- 100% overhead
- + –
cloud: n+1- google: 10% overhead for redundancy
- 3 locations
- + –
- + –
customers moved from office to devices- sitting on 3g dongle because it is faster than local connection
- + –
private cloud -> colocation Facility- ethernet exchange (connection to telco)
- direct local connection to public cloud
- Tomas Kunz (AtoS)
- + –
challanges- requirement of dedication (private cloud)
- allocated resource rather than pay as you go
- on promisse integration with public cloud
- rapid elasticity
- expectation is that everything goes automatically/self service – that is difference from legacy IT
- start with provide scenarios which you want to provide
- + –
MOA/ROA- future of cloud computing and service integration
- Resource oriented architecture
- + –
Security- IAM federation is a must
- Azure – clever: one solution to interact with (SSO)
- what if C level loses his toy?
- + –
fast food / healthy food- expensive
- + –
don’t have- knowladge
- mood
- time
- + –
convenience is megatrend- cloud computing is megatrend – why CIO ignoring it?
- public cloud is more cost effective
- + –
risk- + –
small probabality of plain crash same as winning lotery- ppl fly anyway – it won’t happen
- ppl do lotery – it can happen…
- + –
- free resources by putting things to cloud, but you can use those for innovation!
- providers oversell – gambling with trust
- + –
cloud on premise?- small services – 70%
- enterprise – 20%
- driven by Sam
- + –
problem-
how would minimalistic API allowing you to migrate from one cloud (doesn’t matter if it is public/privat/IaaS/PaaS) to another
it should be very simple – same way as simple smtp is standard today for transfering data between different mail servers
vision: you have iphone/android phone and on it app to drag&drop migration between 2 vendors – what we would need for that
-
- + –
SMAC: Simple Migration API for Cloud- + –
identifier- + –
http://provide/well-known/scmz- /vm-list
- /vm1
- /vm2
- /app1
- /objectData
- + –
- + –
enumeration- + –
list of resource- /mmc
- + –
capabilities- can be part of HEAD
- + –
- + –
operations (CRUD)- + –
GET- to get data from provider
- + –
PUT- to upload data to provider
- large blobs
- some kind of sparating to smaller files during transfare needed
- DELETE (?)
- + –
MOVE/COPY (?)- currently part of WebDAV standard – can we overload those keywords?
- if format not supported then 4xx code
- + –
- + –
- + –
cloudlet: cloud workload format/archive- + –
container- ZIP (optional)
- directory structure
- + –
descriptor (JSON or flat)- + –
Name/Description- human readable
- unicode
- + –
Identifier- UUID or something globaly unique
- icon
- + –
content type- mime
- format of data (VMI, XEN image, WAR, SQL dump etc.)
- or separated to Manifest (list all files and their format)
- + –
- + –
v2 ideas- Manifest
- localization
- signatures
- dependencies
- taxonomy
- hierarchy
- AAA
- + –