Issues with the Latest Version of the Hotel Reservation Benchmark in Kubernetes

[dufanrong]

Hi, thank you for providing such a valuable benchmark for the research. I deployed the latest version of the hotel reservation benchmark in a Kubernetes cluster, but I encountered a few issues:

1. It only supports a very low peak load. Using the same server with 40 CPUs and 128GB of memory, we were able to achieve a QPS of 3000 with the older version. However, with the latest version, the QPS is limited to only 200.

2. We’ve observed a significant increase in tail latency, with the 99th percentile latency reaching 10.83s when the QPS is 300.

3. The newly added microservices do not come with corresponding K8s deployment files. How can we interact with them, such as sending requests?

4. Additionally, I found an issue with Jaeger. As others have mentioned, the latest version of the jaeger-all-in-one does not work properly. When using version 1.61.0, I noticed that most of the requests were not captured by Jaeger. For example, during a load test with wrk at a QPS of 100 for 300 seconds, I only received 293 traces from the Jaeger API, which is less than 1% of the actual query count. In the older version of the hotel reservation benchmark, Jaeger v1.61.0 worked as expected.

We would appreciate it if you could provide the necessary deployment files. These issues are hindering our experiments, and we would be grateful for your assistance.