From 7243bf4de72f4b795f08f6abe99e05da4c06491b Mon Sep 17 00:00:00 2001 From: "Sridhar K. N. Rao" Date: Tue, 28 Nov 2017 16:23:01 +0530 Subject: llc-management: Support for LLC management with RMD This patch adds support for LLC-Last level cache management using RMD. The changes include: 1. 08_llcmanagement.conf: The configuration file to define cache allocation policy. 2. testcase.py: To trigger llc-allocation and cleanup before and after the test, respectively. 3. llc_management/rmd.py: The main file the performs the llc allocation and cleanup. 4. llc_management/resthttp.py: Generic utility to call rest APIs, with http. 5. Fixed a build error due to change in name mismatch. 6. Fixed pylint errors. Override built-in function error. 7. Added empty __init__.py to avoid import errors 8. Fixed deprecated functions errors 9. Fixed copyright issues, removed python-3 checking. 10. Removed pylint disables. 11. resthttp.py is already part of stcrestclient used in tools/pkt_gen/testcenter/ scripts. Hence, deleted from the source. 12. Year update from the license 13. Fixed some testcases.py collision issue JIRA: VSPERF-544 Change-Id: I7cbd155dd66f5a0cef544751841e71b95c9b6821 Signed-off-by: Sridhar K. N. Rao --- tools/llc_management/__init__.py | 17 ++++ tools/llc_management/rmd.py | 198 +++++++++++++++++++++++++++++++++++++++ 2 files changed, 215 insertions(+) create mode 100644 tools/llc_management/__init__.py create mode 100644 tools/llc_management/rmd.py (limited to 'tools') diff --git a/tools/llc_management/__init__.py b/tools/llc_management/__init__.py new file mode 100644 index 00000000..4774dc93 --- /dev/null +++ b/tools/llc_management/__init__.py @@ -0,0 +1,17 @@ +# Copyright 2017-2018 Spirent Communications. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Wrapper for RMD to perform LLC-Management +""" diff --git a/tools/llc_management/rmd.py b/tools/llc_management/rmd.py new file mode 100644 index 00000000..308dda3c --- /dev/null +++ b/tools/llc_management/rmd.py @@ -0,0 +1,198 @@ +# Copyright 2017-2018 Spirent Communications. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +""" +Perform L3-cache allocations for different workloads- VNFs, PMDs, vSwitch etc. +based on the user-defined policies. This is done using Intel-RMD. +Details about RMD can be found in: https://github.com/intel/rmd +""" + + +import itertools +import json +import logging +import math +import socket + +from collections import defaultdict +from stcrestclient import resthttp +from conf import settings as S + +DEFAULT_PORT = 8888 +DEFAULT_SERVER = '127.0.0.1' +DEFAULT_VERSION = 'v1' + + +def cpumask2coreids(mask): + """ + Convert CPU mask in hex-string to list of core-IDs + """ + intmask = int(mask, 16) + i = 1 + coreids = [] + while i <= intmask: + if i & intmask: + coreids.append(str(math.frexp(i)[1] - 1)) + i = i << 1 + return coreids + + +def get_cos(category): + """ + Obtain the Classof service for a particular category + """ + return S.getValue(category.upper() + '_COS') + + +def get_minmax(category): + """ + Obtain the min-max values for a particular category + """ + return S.getValue(category.upper() + '_CA') + + +def guest_vm_settings_expanded(cores): + """ + Check if are running pv+p mode + """ + for core in cores: + if isinstance(core, str) and '#' in core: + return False + return True + + +class IrmdHttp(object): + """ + Intel RMD ReST API wrapper object + """ + + def __init__(self, server=None, port=None, api_version=None): + if not port: + server = DEFAULT_SERVER + if not port: + port = DEFAULT_PORT + if not api_version: + api_version = DEFAULT_VERSION + url = resthttp.RestHttp.url('http', server, port, api_version) + rest = resthttp.RestHttp(url, None, None, False, True) + try: + rest.get_request('workloads') + except (socket.error, resthttp.ConnectionError, + resthttp.RestHttpError): + raise RuntimeError('Cannot connect to RMD server: %s:%s' % + (server, port)) + self._rest = rest + self.workloadids = [] + self._logger = logging.getLogger(__name__) + + def setup_cacheways(self, affinity_map): + """ + Sets up the cacheways using RMD apis. + """ + for cos_cat in affinity_map: + if S.getValue('POLICY_TYPE') == 'COS': + params = {'core_ids': affinity_map[cos_cat], + 'policy': get_cos(cos_cat)} + else: + minmax = get_minmax(cos_cat) + if len(minmax) < 2: + return + params = {'core_ids': affinity_map[cos_cat], + 'min_cache': minmax[0], + 'max_cache': minmax[1]} + try: + _, data = self._rest.post_request('workloads', None, + params) + if 'id' in data: + wl_id = data['id'] + self.workloadids.append(wl_id) + + except resthttp.RestHttpError as exp: + if str(exp).find('already exists') >= 0: + raise RuntimeError("The cacheway already exist") + else: + raise RuntimeError('Failed to connect: ' + str(exp)) + + def reset_all_cacheways(self): + """ + Resets the cacheways + """ + try: + for wl_id in self.workloadids: + self._rest.delete_request('workloads', str(wl_id)) + except resthttp.RestHttpError as ecp: + raise RuntimeError('Failed to connect: ' + str(ecp)) + + def log_allocations(self): + """ + Log the current cacheway settings. + """ + try: + _, data = self._rest.get_request('workloads') + self._logger.info("Current Allocations: %s", + json.dumps(data, indent=4, sort_keys=True)) + except resthttp.RestHttpError as ecp: + raise RuntimeError('Failed to connect: ' + str(ecp)) + + +class CacheAllocator(object): + """ + This class exposes APIs for VSPERF to perform + Cache-allocation management operations. + """ + + def __init__(self): + port = S.getValue('RMD_PORT') + api_version = S.getValue('RMD_API_VERSION') + server_ip = S.getValue('RMD_SERVER_IP') + self.irmd_manager = IrmdHttp(str(server_ip), str(port), + str(api_version)) + + def setup_llc_allocation(self): + """ + Wrapper for settingup cacheways + """ + cpumap = defaultdict(list) + vswitchmask = S.getValue('VSWITCHD_DPDK_CONFIG')['dpdk-lcore-mask'] + vnfcores = list(itertools.chain.from_iterable( + S.getValue('GUEST_CORE_BINDING'))) + if not guest_vm_settings_expanded(vnfcores): + vnfcores = None + nncores = None + if S.getValue('LOADGEN') == 'StressorVM': + nncores = list(itertools.chain.from_iterable( + S.getValue('NN_CORE_BINDING'))) + pmdcores = cpumask2coreids(S.getValue('VSWITCH_PMD_CPU_MASK')) + vswitchcores = cpumask2coreids(vswitchmask) + if vswitchcores: + cpumap['vswitch'] = vswitchcores + if vnfcores: + cpumap['vnf'] = vnfcores + if pmdcores: + cpumap['pmd'] = pmdcores + if nncores: + cpumap['noisevm'] = nncores + self.irmd_manager.setup_cacheways(cpumap) + + def cleanup_llc_allocation(self): + """ + Wrapper for cacheway cleanup + """ + self.irmd_manager.reset_all_cacheways() + + def log_allocations(self): + """ + Wrapper for logging cacheway allocations + """ + self.irmd_manager.log_allocations() -- cgit 1.2.3-korg