Switch to using volatile tables for benchmarking

sky-bench/src/benchtool.rs

@@ -135,13 +135,10 @@ pub fn runner(
 
     let temp_table = libstress::utils::rand_alphastring(10, &mut rand);
     let create_table = libsky::into_raw_query(&format!(
-        "create table {} keymap(binstr,binstr)",
-        &temp_table
-    ));
-    let switch_table = libsky::into_raw_query(&format!(
-        "use default:{} keymap(binstr,binstr)",
+        "create table {} keymap(binstr,binstr) volatile",
         &temp_table
     ));
+    let switch_table = libsky::into_raw_query(&format!("use default:{}", &temp_table));
     let mut create_table_connection = TcpStream::connect(&host).unwrap();
 
     // an okay response code size: `*1\n!1\n0\n`:
@@ -168,8 +165,6 @@ pub fn runner(
         true,
     );
 
-    let drop_table = libsky::into_raw_query(&format!("drop table {}", &temp_table));
-
     // create table
     create_table_connection.write_all(&create_table).unwrap();
     let mut v = vec![0; response_okay_size];
@@ -247,8 +242,9 @@ pub fn runner(
         }
 
         // drop table
+        let flushdb = libsky::into_raw_query(&format!("flushdb default:{}", &temp_table));
         let drop_pool = pool_config.get_pool_with_workers(1);
-        drop_pool.execute(drop_table.clone());
+        drop_pool.execute(flushdb);
         drop(drop_pool);
         dt.iter()
             .for_each(|(name, timer)| report.insert(name, timer.time_in_nanos().unwrap()));