“One of the things that gets me excited about cloud computing is the access to resources and processing capabilities for very large data sets,” – Thomas Lockney, one of CloudCamp Portland’s earliest supporters.
CloudCamp was held June 30th, 2009 from 5:30-10:30 Pm on the 16th floor of WebTrends in Downtown Portland. The unconference was set up for people who work with cloud computing, were interested in learning more, or who wanted to understand what Cloud Computing was all about. You can see some of what was said on Twitter about #cloudcamp, or #cloudcampdx.
This was a very interesting conference that dealt seriously with some very important issues. Many of us in the field will be running into these problems, or already do. The advantages and disadvantages of Cloud computing need to be recognized before they can be dealt with. In this atmosphere (not to mention the excellent weather and balcony we had) information and knowledge sharing seemed to prosper.
The conference began with socializing and then an Un-Panel composed of a handful of campers who were heavily involved in Cloud Computing, either in knowledge or participation. Then, the audience posed a series of questions which were written onto a white board. The panel gave 1-5 minute responses on the questions of their choosing. At the end of the responses and follow up questions, the Dave Nielsen asked how many people were interested in discussion the questions further in an Unconference format. The topics with the most interest became proposed Unconference topics.
This was a unique way to run an Unconfernece. It put everyone on the same page by giving background and preliminary Q+A around key topics. It also allowed experts to distribute knowledge before sessions, and it made it so that everyone got some form of information, so there was less of a liability in missing conference sessions later.
A shout-out to Mr. Walsh, whom I wish I had more time to speak with.
Software as a Service (SaaS). Infrastructure as a Service (IaaS), Platform as a Service (PaaS).
1. Would database as a service be also considering cloud computing?
Mark Johnson: It really depends if you’re an object guy or a relational guy. If you’re a relational guy you might think of it as a platform. If you have a really good database layer, it would be a infrastructure. If you have a business object later it would be a platform.
Dave Nielsen: There are still people who will offer SQL databases as a service, but there’s another type where people just need to store data and store it quick, not necessarily structured, and then there’s a third type where people need to store relational data like SimpleDB.
Right Scale: Your application needs to have a database because it needs to something, or you have some bit Oracle cluster and the application is the database.
Dave Nielsen: Data in the cloud was probably that most popular topic at CloudCamp San Francisco.
here, most of the audience was interested in Data in the cloud.
2. What are the security threats with company data? Solutions?
Mark Johnson: I think I’m answering a slightly different question, but the whole thing of security is — when they bring in security experts when they bring them in and get their opinion on Cloud Computing, they say “it’s not really our issue”, but I think that with cloud computing, it forces people to think about these things sooner.
Marcus: I work with government institutions.
Dave Nielsen: At cloud camp Paris I got a very specific computing. “How can I make sure my data is never seen by the NSA?”
Audience: Don’t ask that in public.
John Hartman: A project I worked on, it was much more secure in the Cloud vs. physical privacy. Easier to rob your house than to go up in the cloud and put that data back together.
3. Any examples of hardware integration?
I didn’t take any notes here. My apologies. If you have something to add, be sure to add it in the comments below.
4. How do you avoid Cloud Lock-in?
Jason Mauer: Issues with wishing to switch from Amazon to something else. How smooth is this transition? Does data get stuck? With Azure, GoDaddy could run a verison of Azure in the CLoud and there would be no issues.And I think we’ll see mroe and more vendors running certain flavors of cloud as Cloud COmputing becomres more prevalent. But I think we’re still in the infancy of cloud computing.
BrowserMob: Google provides a very specific way of turning your data to CLoud. But you have to be careful becase if you write your code to assume that certian pieces will be there, then you can be locked in. Just be careful with it.
Dave Nielsen: If you are interested in security, there’s actually a Cloud Security Alliance. Cloudsecurityalliance.org, contact Nils Puhlmann.
About half the audience was interested in security.
5. What are you running?
Dave Nielsen: how many of you are running something right now?
A third of audience raised hands.
The entire room said Linux.
Debian, Ubuntu, most pop. choices.
(in the cloud?).
6. What are the potential players right now? What do they bring to the table right now?
Dave Nielsen: Just shout them out.
VMWare, Amazon, Ubuntu, SUn wishes they were, Rackspace, possibly Google, Appengine. Some are software providers, but others are Infrastrucre as a Service. If looking at IaaS specifically, GoGrid, Flexiscale, Joyant, Engineyard is insutry – based on top of Ec2 Amazon.
BrowserMob: A small compnay called COntigex that’s rolling out their stuff any day.
Dave Nielsen: BlueLock is a VMware cloud.
7. What are the regulatory compliance issues?
HIPPA, PCI (payment card industry).
8. Are there open source cloud solutions? Cloud as a service?
Right Scale: Yes, out of UC Santa Barbara, they have a program called Eucalyptus which is very similar to Amazon EC2, and it works just like it…for the moment.
Dave Nielsen: Abiquo out of Barcelona (recently moved to SF), also 3tera.
Ed Borasky: Ubuntu, by Canonical out of the UK Intrepid Ibex contains Eucalyptus. They also have something called Nebuli, which I’m not sure what is.
Audience: That’s not part of Ubuntu, but it’s another open source project looking to build another EC2 layer like Amazon.
9. When would you AVOID cloud computing?
Sid (from Jive): When considering enterprise Dave Nielsening, which is very expensive. A lot of problems with some clients where the data can’t leave the warehouse. Also, it’s alittle more expensive because with Cloud Computing you are paying a little bit more for flexibility.
10. How soon will we be talking about connection speeds to the Cloud?
See 13. Performance Issues (question posed by Ed Borasky).
11. What’s the baseline for cloud computing? (When would you move to the cloud?)
Sid: The lead time to to get ne hardware set up can sometimes b 3-4 weeks, but we have a lot of people wh
So sometimes you can run into complicated capacity planning here, where you guess how many people will use it in the next month and then plan it beforehand.
Red Shirt: One way you can use the cloud if you have predictable spiky load, you can use the Cloud to cover it.
Dave Nielsen: Super easy example would be file storage – for images on your website to push them out tho the edge.
Reid Beels: Seems like they’re talking about finished applications. Where would the development process move from local to the Cloud.
Dave Nielsen: At what point did you in the audience move from local to the cloud?
Audience: When the client wanted to see it.
Audience: It actually was when I was steady to deploy.
@dodeja: One instance I saw was with Animoto, with these massive spikes of access. When you’re doing heavy computing it makes sense to push it out onto the cloud.
Dave Nielsen: David Chappell (writes lots of books) – talked about two high uses of cloud, one when you need to scale, and another behind the scenes.
About 5 poeple were interested in use cases of when to move out onto the cloud .
12. IAAS sems to be a popular choice. Amazon seems to be the only one in the game right now. Why does it continue to be the most popular choice?
Makes more sense to Dave Nielsen there.
13. How does an application running in the cloud get accurate performance metrics (Ed Borasky).
BrowserMob: How do you deal with application performance in the cloud? That’s something people have a lot of concern themselves about because all sorts of things, including network bandwidth is not guaranteed. If you’re expecting to get x megabits of upload speed all the time, then that’s not a good mindset. To have the idea when you go in that you don’t know what upload speed there’s going to be is a better idea. If you need better performance, go with the more powerful equipment.
@dodeja: I think it would be more interesting to know the sorts of optimizations you can do to your infrastructure to make it run more smoothly.
Dave Nielsen: but that’s too specific.
Transition to Unconference Planning
Dave Nielsen: We’ll move now into the Unconference part, in which we’ll have 2 sessions of four topics each.
Proposed Unconference Topics
Pricing for different levels of the cloud, different needs.
Say you made a decision to go to the Cloud, but you want to estimate the baseline costs, the spike costs.
Eric was interested in practical approaches to data security for individuals and enterprise level. About half people attended were interested in this.
Practical uses of Amazon. Best practices.
Scott: Deploying Ruby apps in the cloud and making them scream.
Monitoring applications in the cloud.
Adam: Automation system for servers.
Steven Walling: Is Cloud computing a return to time-share mainframe style computing that we were formerly used to? And if so, does that
Lief: was interested in portability in platforms, standards and portability.
Alex Williams: Interested in defining different types of clouds: public clouds, private clouds, hybrid clouds, and use cases for each.
I went to the session on practical approaches to data security for individuals and enterprise level. About half people attended were interested in this.
Eric: It’s not that your data belongs to you – all of your data belongs to us. These larger companies that hold data. I’ve been working on a completely text based data store, flat files. Ideally, I’d like to have everything as secure as possible.
Lets start by defining things that are nice about the Cloud? What’s nice about Software as a Service (SaaS)?
Drew: It’s just easier.
One is reliability and universal access. The availability is everywhere.
Audience: Until a company goes out of business and the data no longer is there.
Aaron Blew: Scale.
Laura F.: Access.
Caseorganic: The fact that you can have one file, accessible by multiple users centrally updated, instead of 6 files, accessible by one person.
Eric: How can we get some of those benefits while still retaining our ownership of that data in the Cloud?
Eric: Academics utilize primitive version control when they keep renaming files over and over, but they often store multiple copies on one hard drive instead of E-mail, and other storage spaces. What I’m suggesting is having a flattened data store that is diversified.
(At this point, I felt like data was becoming a grain store, and that data store needed to be safe from rats and decay so that it would store tons of grain without bursting or being susceptible to storms (data spikes)).
Group on Amazon Tips and Tricks
I arrived at the group after they’d talked about large scale, heavy duty, and enterprise-level storage techniques.
Group host: For the data hobbyist, you can store all of your data on EBS – a data block. Attach it to an individual EC2 instance. You can at least do things like snapshots of it.
Audience: Klint would know something about this, especially EBS.
Klint Finley: We’ve seen big fluctuations with EBS performance. We’ve turned on CloudWatch to kind of see what’s going on.
Dave Nielsen: Do you have a recommended architecture at this point?
Kint: For now we’re trying to do more in memory. Also, caching everything so we can handle spikes in access.
(And during this session I was looking around, thinking, “this is the underbelly – the equivalent of what the printing press is to printers. What lies beneath. The structure of how things work and what things do”. In other words: the most important thing we can be having a conference about right now).
Steven Walling – Is Cloud computing a return to time-share mainframe style computing that wer wwere formerly used to? And if so, does that mean more centralization
Steven Walling: I’m sure you’ve all heard Kevin Kelly’s talk about what technology wants, that what every device will just be a window to the cloud.
@infovore: That everything is a dumb client, and that all the processing is happening up in the cloud.
Steven Walling: but i think that has some of the similar implications, that everything is running through the cloud, or just some of the really important things.
But if everything is running through the cloud there’s the idea that there doesn’t need to be storage anymore. Once everything is in the cloud, you just need a screen and an interface that, you know, you even touch the cloud with.
That entire vision is one extreme of cloud computing, as in, you don’t own anything, you just get to use the resources that someone provides to you.
That was the original idea of computing, that you’d just need a screen and a keyboard.
Bram Pitoyo: Like Thin Client.
Steven Walling: But that these actual computers were so complex and enormous
The reason we did that in the past was because it was cost convenient, and then we pushed it onto the web.
C: But this stuff – this Cloud computing – we’re doing it voluntarily – because it is easier now to store our things on the cloud and then access them from there.
Steven Walling: And what we’re doing is the same thing as before, just flipped upside-down.
Klint Finley: It wasn’t just a time function. you could have a terminal that was a small as a desk that you could access data from the mainframe with.
Joe: But we no longer have the space to be able to store the entire index of the web on your computer. You rely on Google to do that for you.
Some data is so large that you do need it on the cloud.
That was one of the big things Chris Messina was talking about at Open Source Bridge, that there is a need for those big kinds of supermarkets online that provide these large chunks of data service.
StevenWalling: Timeshare computing – too expensive to do anything but Really important science estuff .perosnal computing – anybody can have accress to it everywhere .Does timeshare cut out non-busienss use cases, does cloud cut out business comm?
Caseorganic: I think if a really important business does something online, it will be somewhat secure. But there is not really a set of standards in place for everyone.
Klint Finley: If we had a mesh wireless network it would work out if one network went down.
Jason Mauer: They did air strikes in Iraq in the gulf war to see if they could take down the Internet, and they couldn’t E-mal was used as a test to withstand attack.
Audience: What would happen is that we’d be able to pull off chunks of the Internet and have them function similarly to other chunks.
Audience: I know that a lot of people use Twitter now, or Facebook. A lot of our data is living on those networks now. There’s where I see a lot of problems. How do you get your facebook stuff out? Where does it go? It’s not even structured in the same way as your other data.
Audience: I started using Twitter and followed two people for a while. Now I follow 200. What happened? There’s too much noise. I don’t think I’m ready to handle that much noise yet. What what if I want to step in time? Filter it out? Listen to only the signals I need to?
Eric: It’s question of network structure. If you’re following 20,000 people.
You’re got a representative of every type, 5 people, totally, like Noah’s ark.
You’ve got a DBA, a marketing person. And you’ve got your neighbors, which are total wild-cards. and members of all these tribes i have. It’s about separating that data.
Lief: Yes, but aside from that issue, there’s another. If social networks are like TVs, there are only a few channels. If the channels are owned by giant organizations, then there’s no room for the next Twitter, or Flickr.
Steven :I don’t agree, because the flip side to that is that the guys in the garage don’t have to know anything about database infrastructure in order to know how to build an application. And that weakens the system if many people begin to use it.
Audience: But people are going to want to keep some private data: like family photos, or whoever knows what photos.
Mike Kaos: Consumers are king. They’re going to vote with their bits, so to speak. They’re not going to keep using a service to host their images with their friends, they’re not going to upload their data, unless it’s reliable.
We went over each of the Unconference topics, gathering summaries from participants of each. Since it was quite late, I did not get to take notes beyond the point.
Overall, the conference was a great success. The panel/Unconference hybrid model was refreshing and informative. I experienced only slight frustration in not being able to clone myself to watch simultaneous conference sessions. But this is usual.
Amber Case is a Cyborg Anthropologist and New Media Consultant from Portland, Oregon. She is interested in Cloud computing for many reasons, especially since she uses Twitter @caseorganic, and stores her collection of over 18,000 photos, screenshots, and research notes on Flickr.