All About Clouds

In spring of 2013, VM Farms Inc. Founder and CEO Hany Fahim was approached to give a lecture to University of Toronto students about what the cloud is and how it works. Check out the presentation to learn a little about the VM Farms outlook on cloud-building.

Transcript

2. All about Clouds By: Hany Fahim, Founder and CEO of VM Farms
3. Background • Worked in the hosting industry for 10 years. • Founded VM Farms in 2009. • Based in Toronto. • VM Farms is a Fully Managed Cloud Hosting provider. • All of our customer data is stored in Canada**.
4. What is it?
5. What is a cloud? • Collection of computational resources (hardware and software) delivered over a network, usually the Internet. • 3 distinct aspects makes something a “cloud”: • Resources are usually metered (pay-per-use). • Resources can be provisioned on-demand. • Resources can be manipulated via API.
6. Examples
7. Life before the Cloud...
8. Shared Hosting • Share same hardware and OS. • Low cost. • HostGator starts at $3.96/mo • Fewer options and less flexibility. • Less performant. Frequent outages. • Security implications.
9. Dedicated Hosting • Leased dedicated hardware (servers, network switches, etc...). • High Cost. • Typical monthly fee starts at $200/mo. • Greater flexibility and options. • Higher performance. • Long contracts (typically 1 year, but can be upwards of 3 years).
10. Co-location • Purchase and host your own hardware. • High Initial Cost (CAPEX). • A good server can cost upwards of $10k. • Monthly fees for power, network, cooling. • Long contracts (typically 1 year, but can be upwards of 3 years).
11. VPS • Best compromise between price and performance. • Utilized primitive incarnations of virtualization. • Pay monthly fee for VPS ($40-$100/mo). • Security implications.
12. Which option do I choose?
13. Something’s missing... Datacentres Power Cooling Racks Physical Security Servers RAM CPUs Disk Network Adapters Fans Networks Switches Routers IPs VLANS Firewalls Bandwidth Operating Systems Distributions Packages Configuration Files Services Patching Your Application ?
14. I just want to deploy code...
15. Hire a Sys Admin • Due to complexity, the only option available was to hire a sys admin that knew how to navigate that world. • Sys Admin would prepare the environment, and deploy your code on your behalf since they know the system best. • This leads to Sys Admin vs. Developer wars.
17. The Glory of the Cloud The Cloud Your Application Cloud API
18. The Cloud... • Abstracts away the complication of running a proper server environment. • Controlled through an API so you can “deploy” infrastructure. • Entire environments can be deployed in minutes.
20. What’s it made of? Anatomy of a cloud
21. SAN Backend Network Fabric SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric
22. SAN Backend Network Fabric SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric
23. Compute Nodes • Physical servers running virtualization** software. • Provided CPU and Memory resources. • Usually don’t store any permanent data (ephemeral). • Connected to multiple networks.
24. Virtualization • Virtualization allows you to run multiple operating systems on a single server. • Virtualization usually refers to the Hypervisor. • Motivation for virtualization - most hardware is under-utilized. • Virtualization increases utilization to get the most out of expensive hardware.
25. Types of Virtualization • Fully Virtualized • Complete simulation of the hardware. • Unmodified Guests. • Partial Virtualization • Partial simulation of hardware components. • Modified Guests. • Para-virtualized • Hardware is not simulated - direct isolated access. • Modified Guests.
26. Hardware Hypervisor VM Full Virtualized Device Drviers Device Drviers VM Device Drviers VM Device Drviers Device Drviers Device Drviers
27. Hardware VM Device Drviers Hypervisor VM Para-Virtualized VM Device Drviers Device Drviers
28. Examples • Xen (OSS, full and para-virtualized guests) • KVM (OSS, fully virtualized guests) • VMware (fully and partially virtualized guests) • Microsoft Hyper-V
 (fully virtualized Windows guests) • Solaris Containers (para-virtualized guests)
29. Xen • Most popular and prevalent hypervisor amongst hosting providers. • Amazon, Linode, Rackspace, VM Farms, IBM, HP all use Xen. • Started as a research project at the University of Cambridge. • First version released in 2003. • Citrix acquired XenSource in 2007.
30. SAN Backend Network Fabric SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric
31. Network Layers • One network sits in front of Compute Nodes to provide Internet access. • One network sits in the back of Compute Nodes to provide access to Storage layer. • Usually Ethernet and Fibre Optic based. • Not uncommon to see 10GigE switches.
32. SAN Backend Network Fabric SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric
33. Storage Layer • Use of Storage Area Networks (SANs), Network Attached Storage (NAS), Direct Attached Storage (DAS). • Usually in the form of specialized commercial hardware with lots and lots of drives. • Typically communicate over the iSCSI protocol. • Very *very* Expensive. • You usually have to buy 2 for redundancy. • Failures can be catastrophic.
34. Examples • NetApp • EMC • HP • Dell • And many more...
35. When Clouds crash A hypothetical situation.
36. SAN SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric Backend Network Fabric
37. SAN SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric Backend Network Fabric
38. SAN SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric Backend Network Fabric
39. SAN SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric Backend Network Fabric
40. SAN Backend Network Fabric SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric
41. SAN Backend Network Fabric SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric
42. SAN Backend Network Fabric SAN Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric
43. True Story. • On April 20, 2011, some parts of Amazon Web Services suffered a major outage. • A portion of volumes utilizing the Elastic Block Store (EBS) service became "stuck" and were unable to fulfill read/write requests. • It took at least five days for service to be fully restored.
44. Humans like sleep.
45. Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric Compute Node Compute Node Compute Node Compute Node
46. Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric Compute Node Compute Node Compute Node Compute Node
47. Compute Node Compute Node Compute Node Compute Node Frontend Network Fabric Compute Node Compute Node Compute Node Compute Node
48. How do you use it?
49. IaaS • Infrastructure-as-a-Service. • Ideal for specialization. • Examples: • Amazon Web Services • Rackspace Cloud • VM Farms • Linode
50. PaaS • Platform-as-a-Service. • Ideal for rapid web development. • Examples: • Heroku • Engine Yard • VM Farms • Google App Engine • Windows Azure
51. API • Most providers offer an API to allow you to use the service programmatically. • Script and automate deployment of resources. • Web Interfaces available.
52. Which to choose? • If you’re building an application based on a common framework (Django, RoR, etc...), pick a PaaS provider. • If you’re building a non-standard application that requires specialized resources (such as GPU rendering), pick an IaaS provider. • Or, just come to VM Farms.
53. What are the downsides?
54. Downsides • Clouds are complex - many moving parts. • Outages are common - expect failures. • Expensive. • Data locality.
55. Chaos Monkey • http://www.codinghorror.com/blog/ 2011/04/working-with-the-chaos- monkey.html • Authored by Jeff Atwood. • Created by Netflix to
 create chaos.
56. Where’s my data?
57. USA Patriot Act
    snippet from Wikipedia • One of the most controversial aspects of the USA PATRIOT Act is in title V, and relates to National Security Letters (NSLs). An NSL is a form of administrative subpoena used by the FBI, and reportedly by other U.S. government agencies including the CIA and the Department of Defense (DoD). It is a demand letter issued to a particular entity or organization to turn over various records and data pertaining to individuals. They require no probable cause or judicial oversight and also contain a gag order, preventing the recipient of the letter from disclosing that the letter was ever issued. • Signed into law by President George W. Bush on October 26, 2001. 45 days after 911.
58. This is scary.
59. BC wants data in BC • On October 21, 2004, the BC government amended their Freedom of Information and Protection of Privacy Act (FOIPPA) and put tough restrictions on the storage, access, and disclosing public sector data. • Since then, it is an offense to store such data out of Canada, and even outside of BC under some circumstances.
60. VM Farms stores all data in Canada
61. Building your own.
62. Build your own • OpenStack is an open source IaaS project. • AMD, Intel, Red Hat, Rackspace, Dell, HP, IBM, Cisco, VMware, Yahoo, NASA are on board. • Started by Rackspace and NASA in July 2010.
63. OpenStack Family • Nova (compute) • Swift (object storage) • Glance (image management) • Quantum (network) • Cinder (block storage)
64. The Future
65. The Future... • A WSJ article cites “cloud computing” market to explode to $241B by 2020 (up from $41B in 2011). • Dec 2010 - Heroku was acquired by Salesforce.com for $212B! • https://aws.amazon.com/solutions/case-studies/ lists 422 major websites using AWS, and that’s just Amazon’s list. • https://wiki.openstack.org/wiki/Contributors/Corporate lists 133 *corporate* contributes, with individual contributors in the thousands. • Data locality will continue to be a big concern for all players (providers, end users, government bodies) as more legislation is passed.
66. Questions? hany@vmfarms.com
 VM Farms - vmfarms.com Twitter: @vmfarms