Hostfrequencyscaling (#140)

* Added Host Frequency Scaling Module and Related Extensions. * buildGraph now looks for in mutations as well as out * buildGraph: also search for in-mutations when building * adding cpuburn module in yaml file * error due to const name changes * Adds websocket * revamping frequency scaling * Error correction * error fixing * fixing none to performance * adding exclude of temp none * mapping * getting back NONE mutation * set logging to 10 * making freq scaler discoverable * adding freqscaler disc * bug fix * debug statement * adding more freq * making discoverable by one module * no msg * reverse the discs * changing freq limits * Removing check * removing long debug messages * remove dup discovery of freq scal * Adding debugging statement * buildGraphStage1 rewrite of buildGraph; also no longer require unique start * adding cpuburn * thermalLog info * making thermal log configable * bug fix * fixing variable names * fixing variable * redeclaration issue * createmode * fixing format issue * add file open init * opening file single time * bug * bug * bug * naming convention * error dec * scalerconfigable plus changing thresholds * bug fix * better edge elimination in stage2 graph build; graph sanity check * buildGraph stage 2 now generates sane, discovery dependency aware graphs * freqdiscovery * time in ns * powersave only * adding schedutil scaler * perf * adjusting the thresholds * reversing back scalers to normal flow * changing normal/high temp thresholds * remove debug print statements * making thresholds config flat * error fixing * PS OVER DURATION * PSENFORCE Bug * bug fix * PSENFORCE * Bug fixing in PS Enforcement * Added time bound and thermal bound scaling * bug fixing * Debug statment * bug fix * debug statement * converting current thermal into decimal * DEBUG statement * Bug fix * bug fix * discovering perf after ps * adding node * changing mutation timeout * Adding ticker for thermal bound scenario * setting the muttaion timeout to "1s" * adding condition for powersave * renaming parameters * bug fix * fixing errors from rebase * fixing weird whitespace issue * general cleanup * removing pipxe-child and some other changes * fixing whitespacing * setting child log level back to 7 Co-authored-by: J. Lowell Wofford <[email protected]> Co-authored-by: Kevin Pelzel <[email protected]>
kraken-hpc · Jan 8, 2020 · 4f8d3ad · 4f8d3ad
1 parent b066ff5
commit 4f8d3ad
Show file tree

Hide file tree

Showing 13 changed files with 1,293 additions and 146 deletions.
diff --git a/config/pipxe.yaml b/config/pipxe.yaml
@@ -15,11 +15,13 @@ extensions:
   - github.com/hpc/kraken/extensions/IPv4
   - github.com/hpc/kraken/extensions/RPi3
   - github.com/hpc/kraken/extensions/HostThermal
+  - github.com/hpc/kraken/extensions/HostFrequencyScaler
 # included modules
 modules:
   - github.com/hpc/kraken/modules/restapi
   - github.com/hpc/kraken/modules/websocket
   - github.com/hpc/kraken/modules/rfpipower
   - github.com/hpc/kraken/modules/pipxe
   - github.com/hpc/kraken/modules/hostthermaldiscovery
+  - github.com/hpc/kraken/modules/hostfrequencyscaling
   - github.com/hpc/kraken/modules/cpuburn
diff --git a/extensions/HostFrequencyScaler/HostFrequencyScaler.go b/extensions/HostFrequencyScaler/HostFrequencyScaler.go
@@ -0,0 +1,32 @@
+package hostfrequencyscaler
+
+import (
+	"github.com/golang/protobuf/proto"
+	"github.com/golang/protobuf/ptypes"
+	"github.com/hpc/kraken/core"
+	pb "github.com/hpc/kraken/extensions/HostFrequencyScaler/proto"
+	"github.com/hpc/kraken/lib"
+)
+
+//go:generate protoc -I ../../core/proto/include -I proto --go_out=plugins=grpc:proto proto/HostFrequencyScaler.proto
+
+/////////////////
+// HostFrequencyScaler Object /
+///////////////
+
+var _ lib.Extension = HostFrequencyScaler{}
+
+type HostFrequencyScaler struct{}
+
+func (HostFrequencyScaler) New() proto.Message {
+	return &pb.HostFrequencyScaler{}
+}
+
+func (r HostFrequencyScaler) Name() string {
+	a, _ := ptypes.MarshalAny(r.New())
+	return a.GetTypeUrl()
+}
+
+func init() {
+	core.Registry.RegisterExtension(HostFrequencyScaler{})
+}
diff --git a/extensions/HostFrequencyScaler/proto/HostFrequencyScaler.pb.go b/extensions/HostFrequencyScaler/proto/HostFrequencyScaler.pb.go
diff --git a/extensions/HostFrequencyScaler/proto/HostFrequencyScaler.proto b/extensions/HostFrequencyScaler/proto/HostFrequencyScaler.proto
@@ -0,0 +1,23 @@
+/* HostFrequencyScaler.proto: describes host specific CPU frquency scaling policy objects
+ *
+ * Author: Ghazanfar Ali <[email protected]>, Kevin Pelzel <[email protected]>;J. Lowell Wofford <[email protected]>
+ *
+ * This software is open source software available under the BSD-3 license.
+ * Copyright (c) 2019, Triad National Security, LLC
+ * See LICENSE file for details.
+ */
+
+syntax = "proto3";
+package proto;
+
+message HostFrequencyScaler{
+    enum ScalerState {
+        NONE = 0;
+        POWER_SAVE = 1;
+        PERFORMANCE = 2;
+        SCHEDUTIL = 3;
+
+    }
+    ScalerState state = 1;
+
+}
diff --git a/extensions/HostThermal/proto/HostThermal.pb.go b/extensions/HostThermal/proto/HostThermal.pb.go
diff --git a/extensions/HostThermal/proto/HostThermal.proto b/extensions/HostThermal/proto/HostThermal.proto
@@ -11,13 +11,12 @@
  package proto;
 
 message HostThermal {
-     enum hostThermalState {
+     enum cpuState {
         CPU_TEMP_NONE = 0;
         CPU_TEMP_NORMAL = 1;
         CPU_TEMP_HIGH = 2;
         CPU_TEMP_CRITICAL = 3;
-        //TEMP_WAIT = 1; // Thermal request sent to node
 
     }
-    hostThermalState state = 1;
+    cpuState state = 1;
 }
diff --git a/modules/cpuburn/cpuburn.go b/modules/cpuburn/cpuburn.go
@@ -124,7 +124,7 @@ func (*CPUBurn) NewConfig() (r proto.Message) {
 		ThermalThrottle:  true,
 		ThermalPoll:      1,
 		ThermalResume:    60000,
-		ThermalCrit:      85000,
+		ThermalCrit:      98000,
 		Workers:          4,
 		WorkersThrottled: 0,
 	}