4 ms·

What's on Azure right now?

14 points by amirmc 16y ago

TomOfTTB 16y ago
Things might get better thanks to the new Azure Appliances (which are only 6 months old). I said it when Azure first started: No one wants an abstract cloud. Because it means they're stuck with that platform. If I'm on EC2 I can easily move to rackspace or my own private server. But with Azure I'd have to rewrite the whole darn program. The Azure Appliance doesn't completely solve that but at least it gives you the ability to move to your own data center. Which in turn should make Azure a little more viable (Though Amazon's stolen Azure's best trick with beanstalk)
lucisferre 16y ago
We've been using Azure for a few months now working on a .NET REST API and data service to deploy to it. In the end however while we are building it for Azure we may not end up using. 1. Mediochre deployment story Honestly, it could be far worse. Azure is deployed by uploading a compiled package via an HTTP(S) REST API. There are PS tools allowing this to be scripted too. In fact, all of Azure can be interacted with via their REST API which is one of the good things they did. Despite this, it is nowhere near as nice as a Heroku deployment or any of the other push deployment PaaS providers. 2. The SDK isn't all that great The SDK (now at ver. 1.3) is yet another example of Microsoft's tendency to write nigh untestable APIs (mostly concrete classes, almost no interfaces, etc.). They also apparently don't keep the docs up to date on the latest REST API changes (it was actually suggested that we could use fiddler with their Silverlight management console to figure it out). 3. Azure storage is neat, but in the end insufficient and SQL Azure is missing key features and is expensive Personally, I don't think it makes a lot of sense to build most cloud based applications on top SQL, but that is for another debate. In the end we've chosen to run Mongo on a worker role for our storage needs. I really can't wrap my head around how Azure tables could be useful, and blob storage is really just I way to upload files for CDN or provide storage for VHDs. 4. It's kind of expensive compared to the competition Not much else to say here, but the only reason I can think of for this is that MS doesn't feel the need to compete with the other PaaS providers, which in turn would imply that the typical developer that utilizes PaaS isn't their target market. As seems typical for MS, they are targeting the less competent enterprise and business markets and trying to sell them the cloud like it's the next SQL server, offering lots of integration points (painful as they are to use) with existing Microsoft technology (e.g., you can actually use local AD servers to authenticate with Azure). Bottom line, AFAICT Azure isn't for us.
Flemlord 16y ago
We didn't do a thorough analysis, but got the impression we had a few months of fiddling and rewriting various components to everything work on Azure. We would have taken a closer look if we'd been starting from scratch, but we're trying to move an existing app into the cloud. Also our Plan-B was great--we went to the Rackspace cloud, where we can deploy our app without having to rewrite anything. We're still testing and haven't deployed to clients yet, but so far so good.

14 points by amirmc 16y ago

Things might get better thanks to the new Azure Appliances (which are only 6 months old). I said it when Azure first started: No one wants an abstract cloud. Because it means they're stuck with that platform. If I'm on EC2 I can easily move to rackspace or my own private server. But with Azure I'd have to rewrite the whole darn program. The Azure Appliance doesn't completely solve that but at least it gives you the ability to move to your own data center. Which in turn should make Azure a little more viable (Though Amazon's stolen Azure's best trick with beanstalk)

lucisferre 16y ago

We've been using Azure for a few months now working on a .NET REST API and data service to deploy to it. In the end however while we are building it for Azure we may not end up using. 1. Mediochre deployment story Honestly, it could be far worse. Azure is deployed by uploading a compiled package via an HTTP(S) REST API. There are PS tools allowing this to be scripted too. In fact, all of Azure can be interacted with via their REST API which is one of the good things they did. Despite this, it is nowhere near as nice as a Heroku deployment or any of the other push deployment PaaS providers. 2. The SDK isn't all that great The SDK (now at ver. 1.3) is yet another example of Microsoft's tendency to write nigh untestable APIs (mostly concrete classes, almost no interfaces, etc.). They also apparently don't keep the docs up to date on the latest REST API changes (it was actually suggested that we could use fiddler with their Silverlight management console to figure it out). 3. Azure storage is neat, but in the end insufficient and SQL Azure is missing key features and is expensive Personally, I don't think it makes a lot of sense to build most cloud based applications on top SQL, but that is for another debate. In the end we've chosen to run Mongo on a worker role for our storage needs. I really can't wrap my head around how Azure tables could be useful, and blob storage is really just I way to upload files for CDN or provide storage for VHDs. 4. It's kind of expensive compared to the competition Not much else to say here, but the only reason I can think of for this is that MS doesn't feel the need to compete with the other PaaS providers, which in turn would imply that the typical developer that utilizes PaaS isn't their target market. As seems typical for MS, they are targeting the less competent enterprise and business markets and trying to sell them the cloud like it's the next SQL server, offering lots of integration points (painful as they are to use) with existing Microsoft technology (e.g., you can actually use local AD servers to authenticate with Azure). Bottom line, AFAICT Azure isn't for us.

Flemlord 16y ago

We didn't do a thorough analysis, but got the impression we had a few months of fiddling and rewriting various components to everything work on Azure. We would have taken a closer look if we'd been starting from scratch, but we're trying to move an existing app into the cloud. Also our Plan-B was great--we went to the Rackspace cloud, where we can deploy our app without having to rewrite anything. We're still testing and haven't deployed to clients yet, but so far so good.