Single application pool vs multiple application pool [ASP.NET Core]

-
There are often debates on whether to deploy ASP.NET applications in a single application pool, or assign a application pool for each application. In case of ASP.NET MVC with .NET Framework, the cons involved in the latter are:

  1. Each application pool has it's own memory pool, hence memory will deplete quickly, especially if you are hosting quite a number of applications in your server.
  2. It would difficult for you to perform updates on just a single application, especially if it requires you to turn off the application pool.
#2 can be solved quite easily with the use of load balancer, since you could divert traffic to other server while you perform maintenance on the current server. However, there is no easy solution to #1 but to increase the memory of your server.

Initially, I thought the same would apply to ASP.NET Core applications. However, we need to remember that IIS is just a reverse proxy in ASP.NET Core application, at least if you choose the out-of-process hosting model. Even if we deploy all our ASP.NET applications into a single application pool, we would not be able to save much memory, since each application will be running on the Kestrel server (dotnet.exe) and IIS would just consume a negligible amount of memory as a proxy.

Hence, it would not make much sense to try to cramp everything into a single application pool, especially when .NET Core 2.2 (and above) offers in-process hosting model, which would make your application faster due to the reduction of overhead. You can easily find the benefits of in-process hosting compared to out-of-process.

Comments

Post a Comment

Popular posts from this blog

ASP.NET Core service locator pattern

-
There are many posts out there which specifically say that we should not use service locator pattern (unless necessary), which we would not be able to determine the dependencies during compile time because the dependencies are hidden. However, I am going to point out an issue in case if you need to use it, in the context of ASP.NET Core projects. I have a class Foo , which will be bind to IFoo , and I will need to configure a named HttpClient for that class via the following extension,   services.AddHttpClient<IFoo, Foo>(); If we do the above, Foo will be treated as an object which is Singleton injected. In case you are not aware, we cannot inject services which has a smaller lifespan than the service itself. Hence, we cannot inject services which are transient or scoped into Foo . Here comes the problem, what should we do? That's where service locator pattern comes to the rescue! There might be other way, but let's say we decide to use the service locator provide
Read more

Caching and generic query using NHibernate [.NET Core]

-
One of the important aspects in API design is to reduce the number of fetch request to database. NHibernate provides 2 levels of caching, with the first level cache persist through a session only, while the 2nd level cache persists across multiple sessions. The second level cache can be implemented by installing the NuGet package of NHibernate.Caches.SysCache . Then, it involves some configuration setup to utilize the cache feature. However, there is an alternative way that exists in .NET Core which we could use to build a simpler cache. The caching that we could use is IMemoryCache . Firstly, we could setup the cache during  ConfigureServices . public void ConfigureServices(IServiceCollection services) { services.AddMemoryCache(); } Then, assuming we want to retrieve all the records from a table called DailySales , and we want to keep the records in the cache for 1 minute, we could do the following. private readonly IMemoryCache cache; public DBHelper(IMemory
Read more

Ways to perform eager loading via NHibernate

-
By default we know that lazy loading is enabled in NHibernate. So, let's say if we decide to keep the default, how do we enable eager loading with NHibernate? There are 2 ways to do that (there are more, but I will mention 2 that you will commonly see): 1.  Query<Entity>().Fetch(x => x.ClassA).Fetch(x => x.ClassB); 2. var tempQuery =  Query<Entity>().Fetch(x => x.ClassA).ToFuture();      Query<Entity>(). Fetch(x => x.ClassB).ToFuture();     var entityList =  tempQuery   .ToList(); Do note that Fetch is used to fetch reference, while FetchMany is used to fetch collection. If you are wondering what is the difference between 1 and 2, it is the use of ToFuture to split the SQL statement to multiple queries. The problem with multiple Fetch in a single LINQ query is that it would result in Cartesian product of the main object, in this case, Entity  . Hence, it is not recommended to use multiple Fetch in a single query, but rather use method 2.
Read more