Yes, this is going to happen faster than you expect.
Why Amazon is better than you. (Werner Vogels at NGN’07)
Buy 10,000 computers at a time.
Spread load around the world, throughout the day, across the year.
Get HA and DR for free.
Obsessed with making IT productive.
Consumption-based cost tracking built in.
Developers devoted to building dynamic capacity management tools.
GOOG even builds its own switches!
(Oh, and Google’s hosted services are free. )
Seriously.
“ I’ve got one word for you, boy. Just one word.”
Appliances. =
The appliance of computing is SaaS.
The evolution of computing (with apologies to Forrester) ISP 1.0 Access to the Internet Cloud Your apps on their dynamic infrastructure SaaS Your users on their Internet app BPaaS Your process in their language ISP 2.0 Access to a server on the Internet Colo Your servers in their cages ASP Your data on their apps on their servers
A spectrum of control Cloud Your apps on their dynamic infrastructure SaaS Your users on their Internet app BPaaS Your process in their language Maximum control: The machine is your playground Do whatever Minimum control: Your data in their forms Use their dropdowns Medium control: Your business logic on their platforms Use their coding language
Consider how to analyze productivity, bad behavior
Helpdesk pitfall
Escalation and support paths unclear.
Define who handles what problems
Decide how to tell who’s to blame
How do you know if it’s you or them?
Ensure you have named contacts at the provider
Set response targets and escalation paths
How fast can you get them to escalate it?
Traffic & usage pitfall
The new application significantly changes load on the network, undermining other systems.
Model the network with outbound traffic
Expect firewall changes
Consider time-of-day usage changes
Upgrade bandwidth early
Mobility pitfall
With widely available web UI, everyone expects it to work anywhere on any device.
Know what devices you support
Decide how to log access
Someone will use an iPhone
Legislation pitfall
Your intellectual property isn’t yours any more.
You will use APIs
Make sure what you build is your property
Consider GPL3
Security audits will happen
Make sure you’re allowed to conduct them
Is the provider compelled to help?
Do you need code inspection?
Performance/availability pitfall
The application is slow, or not reliable.
Agree on performance and availability SLAs
For whom ?
What function ?
From where ?
From what component ?
Will have what performance ?
And what availability ?
In what timeframe ?
Clearly state your recourses
“ Using the $100/mo. subscription.”
“ Bank tellers will be able to log in from North American branches with a host latency of under 3 seconds , and will have 99.95% availability during business hours ”
Accounts and SSO pitfall
Activation & termination.
Complexity increases with multiple providers
Ensure there’s a roadmap for SSO
http://www.flickr.com/photos/joelogon/2122303827/
Data I/O pitfall
If you can’t get your data, you can’t migrate.
Lose important bargaining chip
Grab all data
All records
All relations
All metadata
In a standardized format
Several times a year
Firehose pitfall
Your users are overwhelmed and panic.
Decide what extensions are most popular
Roll them out over time to your organization
Reminds them why you went SaaS
Outsourcing pitfall
Failure to commit to the change in focus leaves you supporting half a process.
Once part of a process is gone, how much can you shed?
Physical I/O (mailing, shipping)
Third-party systems (payment, banking, order entry)
Telephony and messaging (VOIP, dialing)
Get executive and financial sponsorship for this!
Upgrade pitfall
When the SaaS upgrades, you’re dragged along.
Get good warning about upgrades
Ensure training is part of support agreement
If possible, upgrade when you want to
Extensibility pitfall
The app isn’t quite right after all.
Make sure you can extend it
With your own code and their APIs
In their development environment
Can you attach new fields to every data structure?
'Ten questions' is a great thing!
Thank you! 2 years ago