On the 25th April, we had the AWS Summit London at ExCel London…

I’ve attended quite a few times now, so I knew what to expect.

The venue, while very, very big is a pain to get to for me to start with. By tube, its approx 1.5 hours if there are no issues with the tubes (that’s three lines) and the last change to get to the venue is always very busy. While facilities at the venue are nice - lots of places to get food/coffee/tea. But be warned that while there is wifi available, usually its very oversubscribed.

I got there without issues on the tube and managed to get into the queue to pick up my badge. And here, Amazon had some issues printing the badges and queues started to build up.

I had a good chunk of time before the keynote, so went to have a look around. My initial impression was that there were possibly fewer vendors there than last year. Some names from last year didn’t seem present. But on the other side of that, it was much busier than the same time last year.

There were tea/coffee stations around as well as water coolers. Last year there had also been fruit and cereal bars available…Glad I’d brought something to snack on my self!

In years past, there was usually a huge amount of swag to be had - t-shirts, pens, stickers, water bottles, keyrings etc, but the past couple of these I’ve noticed a shift to prize draws. For example in 2019, I only noticed a few stands offering prize draws where you get your info scanned and get an entry into the draw towards the end of the day, but most stands were just handing out t-shirts. But this year, it seemed like 80% of the stands were some sort of raffle. If you want to be noticed at these things, you really need to give something away to get people to visit you. There were quite a few claw machines in operation and a spin the wheel (with the worst “prizes” ever - 50% chance of nothing or post-it notes/bubbles and a star prize of a pair of socks!). Any sort of game with a prize gets a queue going and more people wanting to have a go.


Anyway…the keynote. Delivered by Tanuja Randery and Francessca Vasques was heavily focused on Generative AI…again. Yep, pretty much the same theme as last year. Only this time it wasn’t so much as “look at all these LLM models you can use on AWS!”, but “look at what you can do with GenAI on AWS”. You’ve all heard of GitHub Co-Pilot and Microsoft bringing GenAI into everything (Office, Edge, etc), well Amazon are doing the same - there’s a GenAI helper coming to the AWS console and portal to help you, Amazon Q. Looks useful, but it will come at a cost to use, just like Co-Pilot etc.

There was a promo from Adobe about how GenAI on AWS makes their users more creative and faster. The CEO of Zilch appeared and pretty much gave us a sales pitch for Zilch (yawn). I really didn’t get why he was there. We got an overview of what Zilch does and that it uses AWS… The CIO of TUI Group came on and told us about how TUI had gone big on AWS and scaled up massively over the past four years putting everything into AWS and away from traditional data centres. Much more interesting stuff than the Zlich guy.

As always, the keynote over-ran making the first set of breakout sessions redundant - only the people who you see leaving the keynote before its finished and those that skip it manage to get to the first round of sessions. You’d think that some one at Amazon would realise this is the case and schedule accordingly. The keynote finished at 11:05, breakout sessions were due to start at 11:15 and it took 15 minutes to get out of the auditorium and get to the same level as the other sessions - another 5/10 minutes to then make your way through the crowds to the session and, your either too late to get into the session (all the headsets/seats taken) or you’ve missed a big chunk already.

One thing I found interesting during the keynote was a point made about how on-premise infra/apps were a bottleneck for dev and the cloud was the way to improve your apps. Now, over the past year, I’ve seen a number of stories about companies moving out of the cloud and back to traditional data centres (Basecamp was the first company I remember reading about). Cost is probably the biggest factor here. I mean think about it - in the past you’d have been lucky to have dev/test/production environments on premise, but the cloud makes it so much easier to spend without thought. Now I’ve seen dev’s through up full environments to test a text change on a website (app servers, db, Redis etc) which just seems excessive. Policies need to be in place to prevent dev’s wasting money. At one job, I was asked to tag all resources and figure out why their costs were so high for the past two months. After a bit of digging around, someone had created a VM with a massive amount of memory and cpu to do a task…and then left it up, just incase they needed to use it again. It had been sitting idle costing a couple of hundred pounds a day for two months and been used for a day. Reducing waste by having policies/blueprints in place can produce big cost savings.

Which oddly leads me into one of the breakout sessions - How Ocado Technology reduced cloud costs without stifling innovation. Ocado have a pretty impressive setup with their own platform (Ocado Smart Platform) as well as the robot pickers in their fulfillment centres. So they decided to try and reduce their cloud spend without reducing performance or availability. And looking at the graphs they showed, they did manage to do it. By analysing how well auto scaling was working, they made changes to get it to perform better. All sounded really interesting. Also updating to the latest SKU’s can provide some savings. In terms of re-engineering code, updating and switching libraries produced results for them.

Now that’s a thing I’ve noticed in many places, reluctance to update libraries used in code. Often things are left as is until support ends or major vulnerabilities are found requiring an update. If it ain’t broke, don’t fix it shouldn’t be the way you decide whether or not to spend time and effort to update when new versions become available.

The only other session that I managed to get into was about observability. It was pretty much straight forward stuff about using the right measurements to set the right alerts, combining alerts to reduce the numbers etc. It was interesting, but nothing new from my prospective.

Funnily enough, these sessions were on the same themes of the sessions I’d wanted to go to last year, but want able to get in. I’d have liked to have made it into a couple of other sessions, but the second time you find yourself too late to a session due to the crowds and how the break out areas are arranged…you just give up. Miss a session and its a 45 minute wait till the next and that’s a lot of time to kill once you’ve been through the expo a couple of times and seen everything.

To contrast this with an online conference, you pick your sessions and as long as your free, you can join them and watch. All in the comfort of your own surroundings. Miss a session and usually there is a recording that you can watch when its convenient to you. In person, you’ve got to decide and plan out what your going to. You have to take into account the distance things are from each other, the crowds you’ll have to pass through, will there be seats when you get there and more importantly will the headset work to allow you to hear (breakout sessions happen next to each other with cloth dividers between session, so you can only hear whats being said with the headphones in each area). I prefer an online conference.

My AWS cert will be up for renewal next year, so if I do go again I may visit the training workshops. Otherwise I might just skip next year.