Azure provides a
highly scalable, elastic pool of resources that can help you to create super-performant applications. Azure provides obvious things that can help you with performance, like (auto)scaling. It also provides some not-so-obvious services and features that can help your application to perform well.
This article will reveal some of the best services and features in Azure that can help your application to perform:
The thing that often determines the performance of an application is how fast data gets transferred. Think about a web application. The HTML, CSS and JavaScript for it need to be downloaded to your webbrowser to be displayed. The browsers are fast and are getting faster and faster. The machines that run the browsers are already very powerful. Up- and download-speeds are often fast as well.
But still, we are bound by the speed of light. Data travels at the speed of light through wires under the oceans, above land, through routers, to and from datacenters. And that takes time. This takes more time when data needs to travel over long distances. It’s physics.
The time that data takes to travel over distances is called latency. This is the one thing that we can’t really battle (yet). This really influences the performance of applications. This is because it makes requests take longer. And it does this for each request. To load a single page of a website, the webbrowser sometimes has to perform more than 20 requests. When you have a high latency, all of the 20 requests will take longer, which is really noticeable in your application.
Avoid latency
Try to avoid latency in your application. In Azure you deploy your application, services and data in a datacenter, somewhere in the world. You should take care that all of your services are as geographically close together as they can be, preferably in the same Azure region (like Central US). You can find out which Azure datacenter is closest to you by using the
Azure Speed Test website.
Besides making sure that your services have minimal latency when they talk to each other, you should make sure that there is
minimal latency between your users and your application.
Suppose that you have a web application deployed in West Europe. Someone that uses it from Australia will have a high latency, and therefore, a slow experience.
You can deploy the web application
in another Web App, closer to the user, for instance, in Australia East. But how would you route the user to that Web App? You can do that with
Azure Traffic Manager. This is an Azure Service that acts as a global load-balancer that can load-balance based on geographic performance (which is latency).
You couple the domain name of the web application to Traffic Manager, so that users go to Traffic Manager, instead of the Web App. Traffic Manager than determines to which Web App the user should be routed and routes them there.
Minimizing the work that your application has to do, is always good for performance. If you have static content in your application, like images, video- or audio-files, CSS, JavaScript and so on, you can use
Azure CDN to handle the traffic for that content. Azure CDN is basically a hosting platform for content, that copies the content all over the world, so that it is close to the user. By using it, you unburden your application from having to
process the traffic for those resources and you bring the content
closer to the users, which lowers latency.
To use this, you create an Azure CDN profile and choose a provider (Azure works with Verizon and Akamai). Next, you couple an Azure Storage account to the profile, in which you put all of your static content. The CDN profile will then automatically copy all of this content to many Points Of Presence, all around the world, ensuring that the content is always close to the user. In your application, you change the reference to the static content to point to the Azure CDN Profile. A simple, yet very powerful improvement.
Almost every application
works with data. Usually, this data lives in a database of some sort, like an Azure SQL Database or a Document DB. Reading or writing data from or to a database takes time. This is usually because the database needs to gather data by forming relations or traversing indexes and then read or write the data from or to a disk. Your application is better off if it can avoid calls to the database as much as possible.
In Azure, you can do this by using an
Azure Cache for Redis. This is a cache that can (amongst other things) store data in the form of key/value pairs. The advantage of using this instead of a database is that
Azure Redis Cache is fast! That is because the data in the cache is in memory and usually in the format that you want, so it doesn’t require complex queries to be executed.
Usually, you can’t put all of your data in a cache. You should put data in there that doesn’t change often, like lookup tables. By doing that, you avoid database calls for this data and speed up your application.
When you run a .NET application in an
Azure App Services, it will be recycled at some point. By default, each application will be terminated every
1740 minutes (29 hours), which is called a
periodic restart timeout. Also by default, applications will be recycled when they are
unused for 20 minutes, which is called an
idle timeout.
These are side effects from the IIS implementation within Azure App Services. When one of these things happens, your application is ‘restarted’ and will recompile (using Just In Time (JIT) compilation) whenever it is first used. This means that the first users will have to wait a couple of seconds for your app to load.
To avoid this, you can enable the setting Always On in an App Service. This will keep your application ‘alive’ by occasionally hitting it with a request in the root of the application. You do have to run your application in the Basic pricing tier or higher to use this feature.
