May 2, 2019 at 12:00 am
Comments posted to this topic are about the item Serverless
May 2, 2019 at 7:43 am
The ebb and flow of our website traffic is highly variable. It makes sense for many of the middleware functions to be serverless functions. The excitement is not about scaling up, but about scaling down!
In a traditional data centre we buy hardware based on a 3-5 year life cycle and inevitably we try and guess what traffic will be like in 3-5 years time and then for various regulatory reasons scale for 3x peak load. The risk is that we are way off in our guestimate and have paid a hefty sum for something we never need. We don't worry about the equipment not being powerful enough because in that scenario success is its own business case.
Serverless and cloud takes that all away. We only need to pay for what we are running at any one time. The challenge is to break down our applications so that serverless functions are an appropriate solution and also devising ways to have an automated testing regime for serverless functions.
Beyond serverless we have some folks getting very excited about uni-kernels. This is where an application compiles so that it is its own OS/app. In this world BCP would not sit on an OS that in turn handles all hardware (real or virtual) on the server. BCP would be the bit of the OS needed to read/write a file and talk to SQL Server. The vast majority of what an OS does simply isn't needed by BCP so would not exist. This means that uni-kernel apps are miniscule. You are talking about stuff with the memory footprint of an Atari 2600 game cartridge but running on C21st infrastructure!
May 2, 2019 at 1:43 pm
Unikernals sound interesting, but really, more like a container. There still needs to be some hardware manager to allow multiple processes on the hardware, so this seems more like a virtual OS?
I like the idea of smaller footprints for lots of stuff, as we have gotten tremendous bloat as CPU and memory have grown, frameworks increase,and developers stop paying attention to the amount of fluff in code.
May 2, 2019 at 5:48 pm
I believe that Azure Stack and server-less architecture could make financial sense for an organization with a very large IT enterprise; maybe something like a government or global bank where they already have the infrastructure in place, and they want to avoid server sprawl and maximizing their on-prem data center investments.
"Do not seek to follow in the footsteps of the wise. Instead, seek what they sought." - Matsuo Basho
May 16, 2019 at 8:47 am
Very interesting this helped clarify what serverless was for me. I can definitely see that it could be a useful tool to reduce costs for users. Provided its implemented with the user in mind specifically with the idea of adding value for the user (while also benefiting Microsoft) I can see that it will be a success. For some suppliers it could be weighted too much in favour of increasing sales at the expense of clients the pricing levels will be all important and would probably need to be judged on a per case basis.
cloudydatablog.net
Viewing 5 posts - 1 through 4 (of 4 total)
You must be logged in to reply to this topic. Login to reply