I came back from Microsoft Ignite with a very positive impression. The buzzword of this year's Ignite was "planetary scale", and Microsoft has all rights to brag about this with the current Microsoft Azure proliferation. Just about everything they build, they build for their public cloud and to scale planet-wide right out of gate. One interesting feeling I got as an attendee is that at this time, there's basically no way for any Microsoft shop to escape being sucked into the Microsoft Cloud. It's no longer some distant future – we've been cleverly and completely surrounded by one without even noticing, and the event horizon of this "black hole" has already been crossed. There's no way out - and no excuses left not to start using Microsoft Cloud at least to some extent – as by now, Microsoft seems to have an answer for pretty much every objection. I have to admit, this was a brilliant strategy and execution... very clever indeed - Satya Nadella is definitely the best thing that happened to Microsoft since Bill Gates.
The top 3 things that stood out for me personally are as follows. First is Project Honolulu. I attended a few sessions on this one, and came to the realization that unlike everything I've seen before - be it RSAT, Server Manager, Hyper-V Manager, VMM, SCOM - Honolulu is the first management UI from Microsoft that actually makes me wanna use it. The only concern for me is that its current architecture (agentless with no caching database) means Honolulu probably won't scale well. Anyway, at least for SMB it is going to be a killer management tool for sure. I am extremely happy to see this, because I can tell you now that lack decent management UI was the key feedback from MVPs to Microsoft for the past many years, when Microsoft's obsession with PowerShell was steadily hurting their primary selling point – simple, fully UI-driven user experience.
Second technology that really impressed me was Azure Cosmos DB. What impressed me the most about it that despite being next-gen, cloud-native, planetary-scale distributed database – it does not sacrifice legacy database features. For example, as far as the consistency goes, you basically have a menu with as many as 5 consistency levels to choose from between Strong and Eventual, depending on your application's needs – allowing to balance consistency requirements with latency, availability and read scalability. Likewise, due to being cloud-born database, of course it supports all the new fancy APIs – but guess what, if you want to run good old SQL queries against one, you still can! Very impressive.
Finally, I was pleasantly surprised about Microsoft devoting a big chunk of the keynote to quantum computing. Being a big fan of this topic, I really enjoyed this part – especially when those quantum physicists started throwing their jokes. They would not be as funny if the whole situation was not so awkward – you really had to be there to feel it, but try to imagine the awkwardness of the moment when a bunch of scientists at the stage are laughing really hard at their own joke, with the majority of attendees not getting it. Well, I personally did get most of their jokes, so for me this was by far the most interesting IT keynote I ever attended. On a serious note, Microsoft did unveil the quantum computer prototype that they've apparently been working on for 12 years now, so this was impressive. Admittedly, it felt a bit Apple-style aka "take an existing innovation and make it look new and shiny" because we all know there are at least 2 other vendors out there which are already shipping quantum computers (I talked about those in the earlier digests). So I was more impressed with the announced Visual Studio plug-in and 30-qubit quantum computer emulator that you can run on your desktop – this definitely got me intrigued, so I want to try and find some time to see what is this all about.
On a less futuristic note, one impressive feature I learned about was around the integration of the upcoming built-in ReFS deduplication (which I covered here earlier this year) and mirror-accelerated parity set of Storage Spaces Direct (S2D). This is a very cool concept that basically merges the benefits of inline and post-process dedupe approaches, in theory giving you the best of both worlds. Essentially, with this integration all writes land onto S2D caching tier in mirror configuration at full speed (so there's no I/O impact typically associated with inline dedupe), with the data deduped inline as it is being unloaded to S2D capacity tier in parity configuration for best capacity utilization. Sounds like something that will potentially make an awesome backup target! We shall see.
I have a lot more to share but only 15 min before the digest deadline – plus, this post is already getting too long anyway! So just to finish off and make you smile – here's a great coub that sums up all the current AI buzz nicely, by the way something I experienced at Ignite as well ;) having said that, machine learning on the other hand is real - and I actually do consider Microsoft a big innovator there! I feel they have everything lined up nicely and are doing all the right things to win in this space by leveraging their strongest advantage of owning most of the compute power in the world.
Komentáře
Okomentovat