AWS Cloud Summit
At the beginning of the day I fell out of my bed into the cloud summit. Seeing as I was less than a stone's throw away, it was one of the easiest journeys to a conference I had managed.
We registered with our bar-codes from the original e-mail, though this proved to be somewhat of a problem technically, since the bar codes didn't seem to appear on everyone's e-mail. this caused a bit of a technical glitch which delayed some poor souls, but mine worked fine so I was in fairly uneventfully.
The geographical layout was available a day prior to the summit via the Guidebook app, which I have to admit has been one of the most solid platforms that I have encountered for this sort of thing and beat the Azure backed Eventboard app from last year hands down. AWS obviously put some work in to get this right.
What was interesting about the layout was the location of a labs sections near the entrance, with the keynote area cordoned off with high cloth walls which hid a stage where Werner Vogels, CTO of AWS would introduce the summit. The labs areas where we got to try out some of the infrastructure, including the just released (as of today) S3 storage and RedShift storage platforms to the European market. This area also contained large yellow beanbags for the more bohemian amongst us to use to sit, write, blog, experiment etc. I didn't try them out because knowing me, I'd not be able to get back up again because of a) Sleeping or b) Imitating a turtle on it's back.
 |
| The infamous beanbags! My nemesis :-) |
Upstairs, near the black cloaked keynote area, sponsoring organisations and the supporting acts were plying their trade and demonstrating their wares. in the centre of that area was the AWS exhibit itself, manned by a number of architects of the platform, whoc were showcasing some of their logical reference architectures. I had a useful conversation with Andreas about availability, since this is an area I am interested in and have blogged about before. I pointed to various potential single points of failure and and Andreas managed to explain to me the points of availability and how they have solved some of their high availability problems in the reference architecture, which I could read up on and use as case studies. All very interesting stuff.
 |
| The chaps make last minute checks for our arrival |
I was also lured, stomach first, to the Smart421 exhibit by the temptation of boxes of Smarties stacked pyramid-like on their desk. What was interesting about this chat was that they reminded me that unlike the Azure platform, Amazon has a very robust mechanism for having PCI-DSS compliance. After all, they do this every day for their own business. Indeed, Stephen Schmidt went on to introduce the security model under AWS and in particular, the very inspiring and competent VPC (Virtual Private Cloud) offering which uses CloudHSM as a hardware security management option where the client holds the keys and not Amazon. Indeed should anything happen to the physical media, such as tampering or accidental damage, the keys are wiped immediately. This was something that I heard hide-nor-hare of at the Azure conference and PCI-DSS, to the best of my knowledge hasn't been adequately addressed to the same level by Microsoft and certainly isn't as mature.
A little behind schedule, Dr Werner Vogels opened the summit with a comprehensive (read long) keynote on the AWS now and in the future, including some of the new services that have come about recently (53 in the first quarter of 2013 alone). He introduced a number of customers who have used the AWS platform, including a number of very large names in the UK, such as Shell and as well as smaller start-ups such as Shutl.
 |
| Big Werner Vogals on a very big stage |
Werner outlined the strategy for the company and gave us little insights into the internal workings of Amazon. In particular, how Amazon believes in and practises lean principles at their organisation. The insight into how some of the cloud based operations at other organisations seemed to, in part, back up some of the stereotypes associated with the governance of data, such as don't put most sensitive data online. This didn't seem to be a concern at all points and was certainly not one imposed due to any technical limitation, but it did show that some organisations do still have some concerns surrounding the non-colocation of their data.
Steve Schmidt then took over and introduced a number of key information security objectives. He had a lot to say and there was a lot of detail about the AWS security processes and policies. Very impressive as it happens but given they have DoD clients, this is hardly surprising. One of the implicit things that both Steve and Werner introduced without 'saying' in both their own and case study examples was the need for good governance. Indeed, shell were very explicit about their data governance and as such, cloud should never be considered something that should allow your governance processes to become lax. Indeed, the opposite is true in cloud environments. An important lesson there, especially in VPC environments where the clients manage their own keys.
After lunch, which was very popular (So I started to write this blog post until the queue died down), the breakout sessions began. I stuck myself on the bootstrap sessions mostly, but wanted to head into the Architecting for High Availability thread. My phone's battery was being a bit 'negative' and I worried that I would run out of juice from taking all the pictures that I was taking and sure enough, by the time it got to the HA presentation by Ianni Vanvadelis, I had no more juice to take any pics of relevant slides. As it happens, for the customer case study, presented by Dan Richardson, Director of Engineering at Just-Eat, the slides were not there anyway. It was unfortunate, but he actually did a brilliant job in winging it and still managed to get his message across.
Prior to that however, I attended the Bootstrap tracks of "Your first week with Amazon EC2", "Agility and Cost Savings: Achieving the IT 'X-factor'" and the Keynote tracks of AWS OpsWorks, presenting a deep dive with the OpsWorks environment by Thomas Metschke, which was a very technical example of how to combine Chef, Ruby, Git and Jenkins with OpsWorks to deploy to different CM configs. OpsWorks is an excellent platform for this, I have to say.
I went to the X-factor track and my phone died completely. It was tough going from that point onwards, but obviously means I should have known how long my battery would last. However, not being big on photography (with a face like mine, it isn't something that I do much of ;-). Comparing the costings of cloud versus on premise platforms is something I tend to do a lot of anyway, so there wasn't really much new there. However, there is a TCO tool that has been created by AWS for just this purpose. Also, the different costing models were introduced and made very clear. What was interesting was the use of 'Spot pricing' models, which I finally figured out a use for, especially in non-mission critical/off-line work. Reserved instancing is also something I am going to be looking more into. However, it is good to know that I value things the same way AWS do. Having grown up in a house full of economists and financiers hasn't gone to waste ;-)
What was interesting about what Dan Roger told us about was what he didn't tell us. In that session he drew comparison of On-Demand against reserved instances at various levels. you can see the slide above, but what he also shows is the break even with Heavy, Medium and Light Utilisation Reserved instances with each other. In this example, at 8 months it makes no difference which option you choose. This is on top of the very obvious comparison with 1 or 2 month on-demand services.
I spent a lot of time getting a lot of valuable information from the AWS guys playing piggy in the middle with the vendor space about their platforms. In particular, the AWS reference architectures will provide a lot of very useful information, not necessarily in how I would do it, but in providing patterns of best practise from which to work.
As an architect, I have certainly not had anywhere near as valuable a day as I've had today. So I am quite a happy bunny. Plus, having played around with AWS EC2 already, I am much happier than when I was playing around with Azure. There are a couple of blog posts that I can see spinning off this, especially in the area of highly available systems.
All-in-all, a very productive day :-)
After lunch, which was very popular (So I started to write this blog post until the queue died down), the breakout sessions began. I stuck myself on the bootstrap sessions mostly, but wanted to head into the Architecting for High Availability thread. My phone's battery was being a bit 'negative' and I worried that I would run out of juice from taking all the pictures that I was taking and sure enough, by the time it got to the HA presentation by Ianni Vanvadelis, I had no more juice to take any pics of relevant slides. As it happens, for the customer case study, presented by Dan Richardson, Director of Engineering at Just-Eat, the slides were not there anyway. It was unfortunate, but he actually did a brilliant job in winging it and still managed to get his message across.
Prior to that however, I attended the Bootstrap tracks of "Your first week with Amazon EC2", "Agility and Cost Savings: Achieving the IT 'X-factor'" and the Keynote tracks of AWS OpsWorks, presenting a deep dive with the OpsWorks environment by Thomas Metschke, which was a very technical example of how to combine Chef, Ruby, Git and Jenkins with OpsWorks to deploy to different CM configs. OpsWorks is an excellent platform for this, I have to say.
 |
| Thomas takes us through an OpsWorks deep dive. I really like this tbh! |
I went to the X-factor track and my phone died completely. It was tough going from that point onwards, but obviously means I should have known how long my battery would last. However, not being big on photography (with a face like mine, it isn't something that I do much of ;-). Comparing the costings of cloud versus on premise platforms is something I tend to do a lot of anyway, so there wasn't really much new there. However, there is a TCO tool that has been created by AWS for just this purpose. Also, the different costing models were introduced and made very clear. What was interesting was the use of 'Spot pricing' models, which I finally figured out a use for, especially in non-mission critical/off-line work. Reserved instancing is also something I am going to be looking more into. However, it is good to know that I value things the same way AWS do. Having grown up in a house full of economists and financiers hasn't gone to waste ;-)
 |
| AWS Pricing Models presented to us by Dan Roger |
 |
| Dan shows us the break even with the On-Demand services |
What was interesting about what Dan Roger told us about was what he didn't tell us. In that session he drew comparison of On-Demand against reserved instances at various levels. you can see the slide above, but what he also shows is the break even with Heavy, Medium and Light Utilisation Reserved instances with each other. In this example, at 8 months it makes no difference which option you choose. This is on top of the very obvious comparison with 1 or 2 month on-demand services.
I spent a lot of time getting a lot of valuable information from the AWS guys playing piggy in the middle with the vendor space about their platforms. In particular, the AWS reference architectures will provide a lot of very useful information, not necessarily in how I would do it, but in providing patterns of best practise from which to work.
As an architect, I have certainly not had anywhere near as valuable a day as I've had today. So I am quite a happy bunny. Plus, having played around with AWS EC2 already, I am much happier than when I was playing around with Azure. There are a couple of blog posts that I can see spinning off this, especially in the area of highly available systems.
All-in-all, a very productive day :-)
