Improving .NET Aspire eShop web application using Azure Redis Services

This code repository is based on the .NET eShop and eShopLite projects. The application code was refactored to highlight the usage of Azure Redis, and for the convenience of running performance testing. Thanks to the colleagues who worked on the two reference projects!

Pre-requisites

• Visual Studio 2022
• Visual Studio 2022 Preview
• .NET 9
• Docker Desktop

Running the project in Azure using Azure Managed Redis (Preview) B5

This sample uses two AMR instances. One B5 OSS Clustering mode for session store and performance scenarios, and another Enterprise clustering mode for using RediSearch module for Vector Similarity Search.

1. Download this repository

   git clone https://github.com/CawaMS/Aspire_eShop_redis-WIP.git
   

2. Change directory to the AppHost folder

   cd ./Aspire_eShop_redis-WIP/eShopLite.AppHost
   

3. Use Azure Developer CLI tp deploy to Azure

   azd up
   

4. Recommending to use 'Australia East', so resources like Redis, SQL, and Open AI won't run into quota constrains in your Azure subscriptions.

Running the project in Azure using other Azure Redis service SKUs

To clone a branch use:

git clone -b <branch> https://github.com/CawaMS/Aspire_eShop_redis-WIP.git

The follow branches contains the customized bicep files generated from azd infra synth command. The infra folder can be found in the eShopLite.AppHost folder

• OSS Basic C3 (~5G, $131.4/mon) with Entra ID: azd_BasicC3_EntraID

• OSS Basic C3 (~5G, $131.4/mon) with automatically populated connection strings in key vault: azd_BasicC3_ConnectionString

• Azure Managed Redis (Preview) B5 (~5G, $227.76/mon) with High Availability using automatically populated connection strings in Key Vault: azd_AmrB5-HA_ConnectionString

• Azure Managed Redis (Preview) B5 (~5G, $113.88/mon) without High Availability using automatically populated connection strings in Key Vault: azd_AmrB5-noHA_ConnectionString

• A generic project that takes any user-provided connection string: azure_publishRedisAsConnectionString

  • Example connection string for AMR or Redis Enterprise: [redis-instance-name].eastus.redisenterprise.cache.azure.net:10000,password=[redis-access-key],ssl=True,abortConnect=False
  • Example connection string for Redis OSS: [redis-instance-name].redis.cache.windows.net:6380,password=[redis-access-key],ssl=True,abortConnect=False

• .NET 9 Hybrid Cache preview: dotnet9

Features to demo

Output Cache

The eShop product catalog is using Output Cache for fast access. NOTE that this only works without user logged. Code is at Program.cs line 32 and Products.razor line 7

Entra ID

• The .NET 9 Aspire uses Entra ID for Redis by default. See code at Products/Program.cs and eShopLite.AppHost/Program.cs

• The eShop Redis Cache connections are using Entra ID for secure, passwordless auth. Code example at Program.cs line 35

Shopping Cart

The eShop uses Redis to store volatile data like Shopping Cart. Code is at CartService.cs

Leaderboard

The eShop uses Redis SortedSet to save leaderboard of top liked items. Click on the "like" button at the product overview page for a few items and multiple times. The Top Products page would display the sorted items.

The code for using Redis sortedset to power Leaderboard scenario is at ProductEndpoints.cs PostVote method and ProductEndpoints.cs ProductLeaderboard

The Leaderboard feature can also be demo-ed at scale using the SimulateLikeProducts project:

• (Pre-req): Deploy the eShopLite project to Azure

• Obtain the Products service URL, similar to https://products.[domain-name].northeurope.azurecontainerapps.io

• Open the Store service URL to a browser. Click on the Top Product tab

• Ensure user IS NOT logged in

• Open a command prompt in your demo laptop. Change directory to the SimulateLikeProducts project.

  cd SimulateLikeProducts
  

• Set user secrets configuration, build and run the simulation

  dotnet user-secrets set ProductApiUrl  https://products.[domain-name].northeurope.azurecontainerapps.io
  dotnet build
  dotnet build
  

NOTE: Refresh the browser a few times before the automatic refreshing takes place.

• You can see the liked product numbers updating in real time.

  

Demo Leaderboard with Active Geo replication

For geo distributed applications, Redis Enterprise Actvie Geo Replication feature provides simple and success way for each regional deployment to write to the Redis Cache closest to the client application. Redis instances can be configured in an Active geo replication group to have the data synced in real-time.

To setup Leaderboard scenario for Active Geo Replication:

• Create two Azure Managed Redis (Preview) instances in an active geo replication group.

  • Use Redis Enterprise - Create to create geo-replicated Redis Enterprise Cluster 1

    For example, the properties can be:

        {
        "location": "West US 2",
        "sku": {
            "name": "Balanced_B1",
            }
        }
    

  • Repeat step above to create geo-replicated Redis Enterprise Cluster 2

  • Use Databases - Create to create two linked databases in the clusters above.

    For example, the body can be:

    {
        "properties": {
        "clientProtocol": "Encrypted",
        "clusteringPolicy": "EnterpriseCluster",
        "evictionPolicy": "NoEviction",
        "port": 10000,
        "geoReplication": {
        "groupNickname": "groupName",
        "linkedDatabases": [
            {
                "id": "/subscriptions/af4d7ddc-4fc2-4ccf-930d-a8b57a1dcea6/resourceGroups/choliema-rg/providers/Microsoft.Cache/RedisEnterprise/choliema-3/databases/default"
            },
            {
                "id": "/subscriptions/af4d7ddc-4fc2-4ccf-930d-a8b57a1dcea6/resourceGroups/choliema-rg/providers/Microsoft.Cache/RedisEnterprise/choliema-4/databases/default"
            }
            ]
        }
    }
    }
    

• Download the azure_publishRedisAsConnectionString branch

• run azd init and azd up, pass in the connection string of one of the Geo replicated Redis Enterprise databases

• Repeat above step. Pass in the connection string of the second replica of the Geo replicated Redis Enterprise databases

• Put two Store applications side-by-side. Liking one of the applications would update both Top Product pages

  

Vector Similarity Search for Product Recommendations

It's a common feature for eShop websites to show similar products from browsing history. This would increase the chances for customers to purchase by providing more options easily. Click into a Product to view details. The similar items on recommended by Vector Similarity Search on the product descriptions. This is using the ada-textembedding 2 model and RediSearch to achive.

• The code is for creating vector embeddings is at product/Data/DescriptionEmbeddings.cs.

• The code is for performance vector similarity search is at Store/Services/SimilarItemsService.cs.

Distributed Cache

The getProductById is an important page where users can add an item to the shopping cart. The code for doing so in .NET 8 is at ProductEndpoints.cs line 32 and for .NET 9 is at ProductEndpoints.cs line L32. Noticing the simplified programming interface for .NET 9 Hybrid Cache!

To test the performance with more product items and users at scale, we will upload test data and create Azure Load Test.

Upload 5000 test data

1. Change directory to the TestData folder

cd Aspire_eShop_redis-WIP/TestData

2. Set user secrets to point to the Products service API and run the application. Uploading 5000 products may take a while

dotnet user-secrets init  
dotnet user-secrets set ProductApiUrl https://products.<domainName>.<region>.azurecontainerapps.io/api/product
dotnet build
dotnet run

Hydrate cache

Skip this step for SQL only environment

1. Change directory to the TestData folder

cd ..
cd PreloadRedis

2. Set user secrets to point to the Products service API and run the application. Hydrate the cache may take a while.

dotnet user-secrets init  
dotnet user-secrets set ProductApiUrl https://products.<domainName>.<region>.azurecontainerapps.io/api/product
dotnet build
dotnet run

Create Azure Loadtest

1. Create an Azure Load Test
2. Create a test following the instructions below.

• Enter test name and description.

• Configure test plan.

  • Enter the getProductById URL as the target URL. For example: https://products.youracadomaname.northeurope.azurecontainerapps.io/api/product/getProductById
  • Enter the Query parameters. Enter id for Name, ${id} for Value

• Upload the parameter file

  • Browse to the Aspire_eShop_redis-WIP/AzureLoadTest folder. Select the input.csv file.
  • Click the Upload button
  • Enter id as variable
  • Check Split CSV evenly between Test engines

• Skip the Parameters tab

• On the Load Tab, select 4 engines, 250 vUsers each, Test duration to be 7 minutes, Ramp up time to be 2 min.

• Skip Monitoring

• Configure Test Criteria so the report contains P95 and P99 latency, which are common metrics to care about.

• After the test is created, it will start the first run

Run test with multiple environments

Repeat the same process for all the environments you want to run.