Before implementing the Magma storage engine across all our customer sites, we need to conduct two types of tests:
Determine which storage engine (Magma or Couchstore) is more efficient when the application performs disk reads.
Evaluate which storage engine (Magma or Couchstore) is more efficient when the application performs memory reads.
For the first scenario, we plan to use pillowfight to perform SET operations on the cluster with “–key-prefix a,” reducing the resident ratio to less than 100 percent. Then, we will further reduce the resident ratio by performing additional SET operations on the cluster with “–key-prefix b.” Subsequently, we will execute GET operations for “–key-prefix a” in order to prompt couchbase to perform a disk read.
We are currently unsure how best to test scenario-2 in order to assess efficiency during memory reads and would appreciate some guidance.
If the resident ratio less than 100 percent implies disk reads, then the resident ratio greater than or equal to 100 percent implies memory reads. And of course the resident ratio cannot exceed 100 percent.
Hello,
Before implementing the Magma storage engine, we need to test its efficiency compared to Couchstore for disk and memory reads. For disk reads, we’ll use pillowfight to perform SET operations with “–key-prefix a” to reduce the resident ratio, then further reduce it with “–key-prefix b” and execute GET operations for “–key-prefix a” to prompt disk reads. We need guidance on how to test memory read efficiency.
Hello,
I consider using benchmark tools like memtier-benchmark or redis-benchmark. Configure these tools to simulate read operations typical to your application’s workload. Measure key performance indicators such as response times and throughput for both storage engines under varying load conditions.