5. The reality is somewhere in between
●
Need to adapt to a diferent way of doing things
●
The complexity has moved
– Further abstracted from the real world
●
Costs need to be watched
– (Don't forget that double XL instance you left running)
– (And that 20TB of disk)
●
Diversity and resiliance in deployment
– It's gonna break in a odd way, plan for it in a way that
will suit your organisation
6. Adapting
●
It's not going to be easy if you try to transpose your
existing infrastructure to cloud hosting. Consider what is
on ofer:
– Load balencers
– Databases
– Network Security
– Backups
7. The Providers (Examples)
●
“Yellow Box”
– Widest, but also most complex ofering
●
“Blue Box”
– Geared towards Windows
●
“Red Box”
– Morphing from one type of provider to another
●
“Multicoloured big number”
– Racing to the bottom
●
Luring you by waving features in your face
8. Hedonistic IT
●
“Pets” vs “Cattle”
●
Live for today! For we shall die at 7PM!
– Infrastructure wasted 50% of the time
– Maybe you have no idea of the popularity?
– Scale down between semesters
●
Diversity and resiliance in deployment
●
Maybe now you can scale horizontally
– One moodle per dept/school?
– New opportunities for on-billing
10. Case 1 – The Over Estimated
●
Catalyst sub-contracted to build system to deliver one-
shot course for corporations
●
Client over-estimated demand
●
Regular managed-hosting
– Very expensive (Could have purchased machines for
less over year)
– Lead time for hardware ~1 month
– Long complicated setup ~ 2 weeks
●
Over-estimated demand
●
Down to two instances on “Yellow Box”
11. Case 2 – The MOOC
●
Catalyst sub-contracted to build community orientated
system with unknown demand
●
Had to scale quickly
●
Continuous integration
– Cloud computing is great for this, create a stack of
VMs in minutes to do testing on and then destroy
when done
●
Widespread utilisation of providers other oferings
– Cloud Storage
– Cloud Database
– Deployment API and tools
12. Case 3 – The Global Entity
●
Catalyst contracted to build internal online LMS
●
Need to utilise global infrastructure
– Outposts all over the globe
●
Needed hosting meeting all those industry standards
– (Insert various silly standards here)
●
ISO This
●
SAS That
●
SOC Me
●
PCI Card
14. THE DANGER - It's real
●
Cloud hosting is terrifying
●
Can't see internals
– Powerlessness when it stops working (And why?!)
●
Unexpected new challenges
– Rate limiting on all sorts of annoying things
●
Disk access (Ie. Per operation)
●
Network throughput / New connections
●
CPU usage (Watch the %steal)
●
Instance start rates
●
Email sending rate
15. THE DANGER – Continues!
●
It's still shared hosting, and the performance can be
variable
– You are efected by your neighbors; further increasing
your powerlessness
●
Regulatory issues
– Where is my data really?
●
“Equivalent protection”
– Can we trust them not to read it / give to the NSA?
●
Burnt by crappy VPS providers in the past
●
SLAs are pretty open ended, the most you are going to
get back is the money you are putting in
16. THE DANGER – … never ends!
●
To the big guys, you are probably a minnow
– Expect to get treated as such
– (But I bet all are looking for a coup in the education
market)
●
Skill loss
– How networks/firewalls actually work
– Install a database
– Manage storage
●
Scare people
– “We've moved it too the cloud”
18. THE GOOD PARTS – Begin!
●
Free you valuble system administrators to administer
– Save time dealing with hardware vendors
●
Ie. 20 hour crappy desktop
●
Impossible diagnostic tools
– Frequently Redhat (Or worse, Windows) only
●
Impossible staf at $HARDWARE_VENDOR
– Upgrade the BIOS/IMM/$RANDOM_FIRMWARE
– “Not running Windows? We can't help..”
●
Expensive maintenance contracts
– They seldom show up within the agreed time anyway,
and make a weasely excuse (See above)
– And you probably feel like you never use it
19. THE GOOD PARTS – Continue
●
Not wasting time / having 0 lead time on hardware frees
you to
– Try new things
– Do and practise upgrades
– Load testing
●
Not having own hosting means
– No need for proximity to hardware (Sysadmins can
stay at home and not engage in hygiene!)
– Swap CapEx for OpEx
– The vendors are serious about hosting, maybe you
have trouble convincing others to be serious?
20.
21. THE GOOD PARTS – ad nauseam
●
Invest time in resiliency, not redundancy
●
Save the planet (Maybe)
– Cloud providers do more with every W than you can
●
(Watt was Scottish)
●
No more room of that junk because 'maybe we need it'
●
Impress people
– “We've moved it to the cloud!”
23. IN SUMMARY
●
Generally suited to Moodle
– Load balancers
– Cloud DB
– Scale up for busy times
– Huge increase in flexibility
●
Except damn NFS!
●
Cloud hosting is not for everyone
– You still need skilled people to make it work for you
●
It's not the same as outsourcing your Moodle
– Some new challenges to replace old ones