What Elasticity Really Means
Moderator:
Ian Rae, CEO, Syntenic
Speakers:
- Jon Beck, SVP Sales and Client Services, OpSource, Inc.
- Scott Clark, Director of Engineering Infrastructure, Broadcom
- Josh Litwin, President and CEO, Memento Press
- Geir Magnusson, Consulting Architect, Platform, Gilt
- Chad Swartz, Senior Manager, IT Operations, Preferred Hotel Group
Finally, some cloud computing practitioners, small business, enterprise, web retail and technology – Memento Press, Preferred Hotel Group, Gilt and Broadcom.
Alistair Croll mentioned that most enterprises are reluctant to go public with their cloud stories right now, it’s early, and many enterprises don’t want to give away secrets. [This happens in every wave. I speak with shy enterprises as well.]
Drivers:
- Small Business, 1 person IT, find SaaS to help run business
- Enterprise, viable option to expanding data center capacity. Looking for solution to leverage other’s experience, economy of scale, in way that accounts for investment. Pay as you use, for peak periods, is compelling idea for them.
- Web Retail, classic 0 to 60 startup story.
- Technology, best investment option for planned growth. Didn’t go with a complete pay-as-you-go plan. Contracted for a fixed pool of resource that they can use in whatever way see fit. The fixed pool has base plan, overage is pay-per-use, have opportunity to adjust plan.
Interesting points regarding Elasticity:
- OpSource, most customers are more concerned with ability to scale-down, rather than scale-up. Don’t want to pay for unused capacity. But, want to know capacity is available.
- Preferred Hotels emphasizes the need to evaluate/change application distribution and infrastructure setup for performance. You might end up using more server instances in the cloud to get the same performance. If so, need to adjust your software licenses accordingly.
- Gilt’s biggest concern is horizontal scale. Were concerned with ‘Oprah effect’ on Gilt infrastructure, needed to ensure scale and performance. Mentions challenges of database scaling. Really need to look at applications and prepare them for scale.
- Broadcom, there is an appeal to participate in something bigger than them. Leverage ecosystem, economy of scale, really smart people. Foresees purpose built clouds, for a business activity.
Recommendations:
- Gilt, despite all the hype, this (cloud) is a fundamental change. Application architecture is changing. Operations management is changing. Need to understand where cloud is going. The hype will end, the changes will stay.
- Preferred Hotels, this is new era. Doesn’t make sense to build-out and manage these computing environments. Starting points: virtualization or web applications. States “make commitment to virtualization”. Calls out agility, time to setup new environment, as key return.
Audience Q&A
Q: How do you know when you need to spin up a new server?
A: Transparency and monitoring – Opsource
A: Looks like ‘regular’ server, can see highs and lows, manage as ‘regular sever’. – Gilt
Q: Plan servers?
A: Have predictable demand, spikes at noon, build up before, drop off after. Base is steady demand, elasticity for peak
A: Over provision individual servers, common configuration, allows for agility and ease of management.
Q: Application changes for performance on cloud?
A: Not the cloud, it’s virtualized environment that pushes need for change. Rebuild servers, not necessarily applications. But, applications (and application architecture) need to be reviewed.
My bottom line interpretation: Scale does not equal performance!
Related posts:
