Updated READMEs for system benchmarks

README.md

@@ -101,6 +101,8 @@ major production workloads listed as follows:
   </tr>
 </table>
 
+### WDLBench
+
 Also, DCPerf includes micro-benchmarks focusing on hot and common functions/routines in the datacenter.
 
 **Note:** WDLBench has different usages than other benchmarks. To use it, please first go
@@ -124,6 +126,34 @@ specific [README](packages/wdl_bench/README.md).
   </tr>
 </table>
 
+### System Bench
+DCPerf's system bench focuses on system performance metrics (kernel scheduler and syscalls).
+
+
+<table>
+  <tr>
+   <td>Benchmarks </td>
+   <td>Programming Languages</td>
+   <td>LIbraries / SW Stack</td>
+   <td>Application domain they represent</td>
+  </tr>
+  <tr>
+   <td>
+    <a href="packages/schbench/README.md">Schbench</a>
+   </td>
+   <td>C</td>
+   <td>Kernel Scheduler Benchmark</td>
+   <td><a href="https://kernel.googlesource.com/pub/scm/linux/kernel/git/mason/schbench">Linux Kernel Schbench</a></td>
+  </tr>
+  <tr>
+   <td>
+    <a href="packages/syscall/README.md">Syscall</a>
+   </td>
+   <td>C++</td>
+   <td>clock_gettime, getpid, nanosleep, tcp-related</td>
+   <td>Common system call benchmarks</td>
+  </tr>
+</table>
 
 ## Representativeness
 

packages/schbench/README.md

@@ -0,0 +1,14 @@
+# Schbench
+
+Benchpress includes Linux kernel's schbench tool.
+
+Schbench can be invoked using benchpress cli:
+
+```
+./benchpress_cli.py -b system run schbench_default
+```
+
+Or, with custom runtime:
+```
+./benchpress_cli.py -b system run schbench_custom -i '{"runtime":20}'
+```

packages/syscall/README.md

@@ -0,0 +1,26 @@
+# Syscall Benchmarks
+
+Benchpress supports several syscall benchmarks. By default, it benchmarks all of them.
+
+Single core benchmark:
+```
+./benchpress_cli.py -b system syscall_single_core
+```
+
+Multi core (use all available cores):
+```
+./benchpress_cli.py -b system syscall_autoscale
+```
+
+The syscall benchmark supports the following arguments:
+
+```
+./benchpress_cli.py -b system syscall_single_core \
+    -i '{"workers": 1, "duration_s": 60, "nanosleep_ns": 100, "base_port": 16500, "syscalls": "clock_gettime,getpid,nanosleep,tcp"}'
+```
+Where:
+- `workers`: Number of threads to use. -1 to use all cores
+- `duration_s`: Test duration in seconds
+- `nanosleep_ns`: Duration of nanosleep in nanoseconds
+- `base_port`: Base port for TCP server
+- `syscalls`: Comma separated list of syscalls to test. All if empty