For some reason an Azure outage does not faze me in the same way that an AWS outage does.
I have never had much confidence in Azure as a cloud provider. The vertical integration of all the things for a Microsoft shop was initially very compelling. I was ready to fight that battle. But, this fantasy was quickly ruined by poor execution on Microsoft's part. They were able to convince me to move back to AWS by simply making it difficult to provision compute resources. Their quota system & availability issues are a nightmare to deal with compared to EC2.
At this point I'd rather use GCP over Azure and I have zero seconds of experience with it. The number of things Microsoft gets right in 2025 can be counted single-handedly. The things they do get right are quite good, but everything else tends to be extremely awful.
Many years back was the first time I used Azure, evaluating it for a client.
I remember I at one point had expanded enough menus that it covered the entirety of the screen.
Never before have I felt so lost in a cloud product.
The "Blades" experience [0] where instead of navigating between pages it just kept opening things to the side and expanding horizontally?
Yeah, that had some fun ideas but was way more confusing than it needed to be. But also that was quite a few years back now. The Portal ditched that experience relatively quickly. Just long enough to leave a lot of awful first impressions, but not long enough for it to be much more than a distant memory at this point, several redesigns later.
[0] The name "Blades" for that came from the early years of the Xbox 360, maybe not the best UX to emulate for a complex control panel/portal.
Azure to me has always suffered from a belief that “UI innovations can solve UX complexity if you just try hard enough.”
Like, AWS, and GCP to a lesser extent, has a principled approach where simple click-ops goals are simple. You can access the richer metadata/IAM object model at any time, but the wizards you see are dumb enough to make easy things easy.
With Azure, those blades allow tremendously complex “you need to build an X Container and a Container Bucket to be able to add an X” flows to coexist on the same page. While this exposes the true complexity, and looks cool/works well for power users, it is exceedingly unintuitive. Inline documentation doesn’t solve this problem.
I sometimes wonder if this is by design: like QuickBooks, there’s an entire economy of consultants who need to be Certified and thus will promote your product for their own benefit! Making the interface friendly to them and daunting to mere mortals is a feature, not a bug.
But in Azure’s case it’s hard to tell how much this is intentional.
I still feel lost just trying to view my application logs.
I don't want to pay for or lock myself into, "Azure Insights".
I just want to see the logging, that I know if I can remember the right buttons to click, are available.
The worst place to try is "Monitoring > Logs", this is where you get faced up front with a query designer. I've never worked out how to do a simple "list by time" on that query designer, but it doesn't matter, because if you suffer through that UX, you find out that's not actually where the logs are anyway.
You have to go down a different path. Don't be distracted by "Log Stream", that's not it either, it sounds useful but it's not. By default it doesn't log anything. If you do configure it to log, then it still doesn't actually log everything.
What you have to actually do, and I've had to open the portal to check this, is click "Diagnose and Solve Problems" and then look for "Diagnostic tools" and then a small link to "Application Event Logs".
Finally you get to your logs, although it's still a bad way to try to view logs, it's at least marginally better than the real windows event viewer, an application that feels like it hasn't been updated since NT4. ( Although some might suggest that's a good thing. )
A lot of Azure Insights is a value add on top of OTel. You can use "just" OTel feeding Azure Insights (and that's what the modern Aspire-influenced defaults mostly do) and possibly avoid that feeling of vendor lockin. That perspective might also give you ideas of other "modern" vendors to audition such as Grafana if you wanted to see what other people are doing with OTel rather than Event Logs and file system logs.
I love that OTel exists, but I've always been confused why OTel even existed. If I were running a cloud provider, I'd put a "skunkworks" team on all things observability and log management a decade ago, and make their product free to use. It's not critical-path on infrastructure, so it's not a correlated risk other than the investment in manpower, but it dramatically changes how user-friendly your web interface is, and how likely people are to use it daily vs. managing things over command line or with third-party interfaces.
By bringing those eyeballs onto your cloud console, you're creating infinitely more opportunities for branded interaction and discovery of your other cloud products - you could even quantify these eyeballs as you would ad inventory! There should have been an arms race for each cloud provider to have the best log-tailing and log-searching and log-aggregation system imaginable. OTel could have been killed before it began, because Honeycomb and its other originators would have been acquired years ago and made specific locked-in value-adds for each cloud.
But nobody had this foresight, and thus comments like yours are absolutely correct. OTel is a blessing and I love the tools coming out. But from a cloud provider's perspective, it's a massive missed opportunity that continues to be missed.
> By bringing those eyeballs onto your cloud console, you're creating infinitely more opportunities for branded interaction and discovery of your other cloud products - you could even quantify these eyeballs as you would ad inventory! There should have been an arms race for each cloud provider to have the best log-tailing and log-searching and log-aggregation system imaginable. OTel could have been killed before it began, because Honeycomb and its other originators would have been acquired years ago and made specific locked-in value-adds for each cloud.
I think that's what Application Insights has always been, Azure's free-to-start, suggest-out-of-the-box Honeycomb. App Insights had a long slow road away from Microsoft-specific log and metrics ingesters that weren't OTel, but it is hard to argue that standard ingestors are a bad idea. App Insights still downplays that it can be "just a Honeycomb" using only OTel sources and still encourages "secret sauce" ingestors in addition to OTel ones. App Insights is a small moat (around a data lake; to mix metaphors). That said, it's also a standards-supporting tool now as well.
It's not been as clear of an arms race because AWS and GCP didn't invest in it in a similar way and it mostly impacted what are often called "dark matter" teams (Microsoft shops doing "boring" stuff that rarely makes HN headlines), but I have worked in teams that absolutely favored Azure over AWS/GCP with one of the reasons being Application Insights was an easy install and powerful first-party supported tool rather than an extra third party vendor relationship like Grafana/Honeycomb/Dynatrace/etc.
Not sure what to imagine with this given I didn't use Azure at the time. Is this like the Windows XP style task menu?
Think Niri [0], but worse, embedded in a web browser tab, and without keyboard navigation.
Here's a somewhat ancient Stack Overflow screenshot I found: https://i.sstatic.net/yCseI.png
(I think that's from near the transition because it has full "windowing" controls of minimize/maximize/close buttons. I recall a period with only close buttons.)
All that blue space you could keep filling with more "blades" as you clicked on things until the entire page started scrolling horizontally to switch between "blades". Almost everything you could click opened in a new blade rather than in place in the existing blade. (Like having "Open in New Window" as your browser default.)
It was trying to merge the needs of a configurable Dashboard and a "multi-window experience". You could save collections of blades (a bit like Niri workspaces) as named Dashboards. Overall it was somewhere between overkill and underthought.
(Also someone reminded me that many "blades" still somewhat exist in the modern Portal, because, of course, Microsoft backwards compatibility. Some of the pages are just "maximized Blades" and you can accidentally unmaximize them and start horizontally scrolling into new blades.)
[0] https://github.com/YaLTeR/niri
azure likes to open new sections on the same tab / page as opposed to reloading or opening a new page / tab (overlays? modals? I'm lost on graphic terms)
depending on the resource you're accessing, you can get 5+ sections each with their own ui/ux on the same page/tab and it can be confusing to understand where you're at in your resources
if you're having trouble visualizing it, imagine an url where each new level is a different application with its own ui/ux and purpose all on the same webpage
Imagine OG Xbox menus, or the PS3/PSP menus.
Count your blessings. You could have to use Azure SSO through Oracle Cloud..... ; ;
AWS' UI is similarly messy, and to this day. They regularly remove useful data from the UI, or change stuff like the default sort order of database snapshots from last created to initial instance created date.
I never understood why a clear and consistent UI and improved UX isn't more of a priority for the big three cloud providers. Even though you talk mostly via platform SDK's, I would consider better UI especially initially, a good way to bind new customers and pick your platform over others.
I guess with their bottom line they don't need it (or cynically, you don't want to learn and invest in another cloud if you did it once).
It’s more than just the UI itself (which is horrible), it’s the whole thing that is very hostile to new users even if they’re experienced. It’s such an incoherent mess. The UI, the product names, the entire product line itself, with seemingly overlapping or competing products… and now it’s AI this and AI that. If you don’t know exactly what you’re looking for, good luck finding it. It’s like they’re deliberately trying to make things as confusing as possible.
For some reason this applies to all AWS, GCP and Azure. Seems like the result of dozens of acquisitions.
I still find it much easier to just self host than learn cloud and I’ve tried a few times but it just seems overly complex for the sake of complexity. It seems they tie in all their services to jack up charges, eg. I came for S3 but now I’m paying for 5 other things just to get it working.
Any time something is that unintuitive to get started, I automatically assume that if I encounter a problem that I’ll be unable to solve it. That thought alone leads me to bounce every time.
100% agree. I've been working in the industry for almost 20 years, I'm a full stack developer and I manage my servers. I've tried signing up for AWS and I noped out.
AWS Is a complete mess. Everything is obscured behind other products, and they're all named in the most confusing way possible.
I found it intuitive but admittedly it felt a lot like their Xbox UI which I used a lot during my formative years
Amazon: here's two buttons, some check boxes and a random popup.
MSFT : Hold my beer...
> At this point I'd rather use GCP over Azure and I have zero seconds of experience with it.
TBH, GCP is very good! More people should use it.
I know for some people the prospect of losing their Google Cloud access due to an automated terms of service violation on some completely unrelated service is worrisome.
https://cloud.google.com/resource-manager/docs/project-suspe...
I'd hope you can create a Google Cloud account under a completely different email address, but I do as little business with Google as I can get away with, so I have no idea.
That's generally speaking a good practice anyways. My Amazon shopping account has a different email than my Amazon Web Services account. I intuited that they need to be different from the get go.
I haven't used much of GCP, but I have had a good experience with Cloud Run and really haven't found a comparable offering from the other clouds.
Cloud Run is incredible. It’s one of those things I wish more devs knew about. Even at work where we use GCP all the “smart” devs insist on GKE for their “webscale” services that get dozens of requests a second. Dozens!
Isn’t ECS Fargate pretty much the same thing?
ACI is the corresponding service in our favorite cloud.
>I've used AWS, Azure, and recently GCP. You do NOT want to use GCP.
>TBH, GCP is very good! More people should use it.
These takes couldn't be further apart. Gotta love HN comments.
GCP console is not the best but as a long term multicloud user, I can assure you that GCP is much better than Azure portal and/or Azure APIs which is fucking hell
have used both. I prefer Azure. Much easier to use. found GCP unintuitive.
The problem is that in some industries, Microsoft is the only option. Many of these regulated industries are just now transitioning from the data center to the cloud, and they've barely managed to get approval for that with all of the Microsoft history in their organization. AWS or GCloud are complete non-starters.
I moved a 100% MS shop to AWS circa 2015. We ran our DCs on EC2 instances just as if they were on prem. At some point we installed the AAD connector and bridged some stuff to Azure for office/mail/etc., but it was all effectively in AWS. We were selling software to banks so we had a lot of due diligence to suffer. AWS Artifact did much of the heavy lifting for us. We started with Amazon's compliance documentation and provided our own feedback on top where needed.
I feel like compliance is the entire point of using these cloud providers. You get a huge head start. Maintaining something like PCI-DSS when you own the real estate is a much bigger headache than if it's hosted in a provider who is already compliant up through the physical/hardware/networking layers. Getting application-layer checkboxes ticked off is trivial compared to "oops we forgot to hire an armed security team". I just took a look and there are currently 316 certifications and attestations listed under my account.
https://aws.amazon.com/artifact/faq/
I've found that lift and shifting to EC2 is also generally cheaper than the equivalent VMs on Azure.
Microsoft really wants you to use their PaaS offerings, and so things on Azure are priced accordingly. A Microsoft shop just wanting to lift-and-shift, Azure isn't the best choice unless the org has that "nobody ever got fired for buying Microsoft" attitude.
I've used AWS, Azure, and recently GCP. You do NOT want to use GCP.
Microsoft is better at regulatory capture, so Azure has many customers in the public sector. So an Azure outage probably affects the public sector more (see example above about trains).
Microsoft has the regulatory capture. All the European privacy and regulatory laws are good for Azure. That's why your average European government or baking app runs most likely on Azure. (or Oracle, but more likely Azure)
What Amazon, Azure, and Google are showing with their platform crashes amid layoffs, while they supports governments that are Oppressing's Citizens and Ignoring the Law, is that they do not care about anything other than the bottom line.
They think they have the market captured, but I think what their dwindling quality and ethics are really going to drive is adoption of self hosting, distributed computing frameworks. Nerds are the ones who drove adoption of these platforms, and we can eventually end if we put in the work.
Seriously with container technology, and a bit more work / adoption on distributed compute systems and file storage (IPFS,FileCoin) there is a future where we dont have to use big brothers compute platform. Fuck these guys.
These were my thoughts exactly. I may have my tinfoil hat on, but outages these close together between the largest cloud providers amid social unrest, my wonder is the government / tech companies implementing some update that adds additional spyware / blackout functionality.
I really hope this pushes the internet back to how it used to be, self hosted, privacy, anonymity. I truly hope that's where we're headed, but the masses seem to just want to stay comfortable as long as their show is on TV
> they do not care about anything other than the bottom line.
if all companies focused on fixing each and every social issue that exists in the world, how would they make any money?
All I'm saying Its a complete 360 from the marketing they did to establish their brand... you know, save the world kind of stuff, especially from google.
From 2000-2016 most tech marketing\branding was aimed at some kind of social benefit.
Preach
The only reason you'd notice MS was down was if Github was down....
GitHub doesn't use Azure yet, but has just published their migration path to azure a few days ago.
I would link to that article, but that one does seem down ;)
https://www.githubstatus.com/incidents/4jxdz4m769gy
> They're stating they're working with the Azure teams, so I suspect this is related.
At least some bits of it do. I was writing something to pull logs the other day and the redirect was to an azure bucket. It also returned a 401 with the valid temporary authed redirect in the header. I was a bit worried I'd found a massive security hole but it appears after some testing it just returned the wrong status code.
I only noticed it because VSCode's updater failed.
I like azure. I think they are more intuitive tua aws and good, and they have good prices for startups ( essentially free for a whole year )
What did you do when AWS was down last week?
Azure outages don’t faze anyone because nobody notices when it happens.
I read "Microsoft shop" as "Microsoft slop". Fitting. But at least they open source wash themselves so much they're practically a charity right?