Tair (Redis OSS-compatible) instances sit at the data layer, close to your application services, and handle a high volume of requests. Each instance type has a maximum bandwidth limit. When traffic consistently exceeds that limit, application performance degrades.
This page walks you through identifying the cause of high traffic and resolving it.
Step 1: Check traffic usage
Query the traffic usage of the instance for the time period when you suspect the issue occurred. Focus on two metrics:
Inbound traffic utilization (Intranet In Ratio)
Outbound traffic utilization (Intranet Out Ratio)
If either metric averages 80% or higher consistently, the instance is approaching its bandwidth limit and you should investigate. In severe cases — such as the example below — both metrics spike to 100%.
Figure 1. Traffic usage example
If you already know when the spike occurred, skip this step and go directly to Step 2: Identify the cause.
Tip: Set up alert rules on Intranet In Ratio and Intranet Out Ratio at 80% to get notified before traffic reaches the limit. See Alert settings. Acting at 80% gives you time to investigate before service is affected.
Step 2: Identify the cause
Three common causes can drive high traffic on a Tair (Redis OSS-compatible) instance.
Before you start: Temporarily adjust the instance bandwidth to reduce the immediate impact on your services while you work through the root cause.
Check each cause in order.
Large keys and hot keys
Use the Top Key Statistics feature to identify large keys or hot keys on the instance. The feature displays key details directly in the console.
These two key types cause traffic problems in different ways:
| Key type | How it causes high traffic | Resolution |
|---|---|---|
| Large key | Each access transfers a large payload, which can saturate bandwidth and cause traffic spikes. | Split by business logic (e.g., user ID or time range), reduce access frequency, or delete keys that are no longer needed. See Large keys and hot keys. |
| Hot key | High-frequency access to the same key generates a continuous stream of small requests that accumulate into sustained traffic elevation. | Split by business logic, or use the Proxy Query Cache feature to cache hot keys at the proxy layer. |