)]}'
{"specs/victoria/approved/time-series-framework.rst":[{"author":{"_account_id":21692,"name":"licanwei","email":"li.canwei2@zte.com.cn","username":"licanwei"},"change_message_id":"6119b9945a0bba7af15b09ce0dd3c45e172af0af","unresolved":false,"context_lines":[{"line_number":48,"context_line":"Proposed change"},{"line_number":49,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"To enable time series the datasource base class will need a new method. This"},{"line_number":52,"context_line":"method will allow metrics to be retrieved over a specific period with a"},{"line_number":53,"context_line":"specific granularity. This method will return a list of values according to the"},{"line_number":54,"context_line":"supplied parameters and will subsequently be implemented by all current"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_b87bbc89","line":51,"range":{"start_line":51,"start_character":0,"end_line":51,"end_character":70},"updated":"2020-06-26 07:21:55.000000000","message":"Are you sure the datasources such as gnocchi can get a time series?","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":29911,"name":"Dantali0n","email":"info@dantalion.nl","username":"Dantali0n"},"change_message_id":"eef82d6f400c24cd296a8259924585de8e5d7cac","unresolved":false,"context_lines":[{"line_number":48,"context_line":"Proposed change"},{"line_number":49,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"To enable time series the datasource base class will need a new method. This"},{"line_number":52,"context_line":"method will allow metrics to be retrieved over a specific period with a"},{"line_number":53,"context_line":"specific granularity. This method will return a list of values according to the"},{"line_number":54,"context_line":"supplied parameters and will subsequently be implemented by all current"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_f948612f","line":51,"range":{"start_line":51,"start_character":0,"end_line":51,"end_character":70},"in_reply_to":"bf51134e_3ed91f22","updated":"2020-07-01 05:54:12.000000000","message":"There is effectively no new datasource the only thing that changes is that instead of always retrieving one value from ceilometer, gnocchi, grafana etc: a list of values can be retrieved. I will update the spec to reflect how this new method could look.","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":29911,"name":"Dantali0n","email":"info@dantalion.nl","username":"Dantali0n"},"change_message_id":"119827acf2525135656d09ffb8780eb78401b317","unresolved":false,"context_lines":[{"line_number":48,"context_line":"Proposed change"},{"line_number":49,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"To enable time series the datasource base class will need a new method. This"},{"line_number":52,"context_line":"method will allow metrics to be retrieved over a specific period with a"},{"line_number":53,"context_line":"specific granularity. This method will return a list of values according to the"},{"line_number":54,"context_line":"supplied parameters and will subsequently be implemented by all current"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_b8f57c59","line":51,"range":{"start_line":51,"start_character":0,"end_line":51,"end_character":70},"in_reply_to":"bf51134e_b87bbc89","updated":"2020-06-26 08:21:28.000000000","message":"Gnocchi is an interesting case as it stores time-series in a condensed aggregated format. I suspect this won\u0027t necessarily be a problem but it will likely mean Gnocchi its granularity will be worse compared to other time-series datasources.\n\nFor example the openstack cli supports `metric measures aggregation` which returns measurementS of aggregated values. \nhttps://docs.openstack.org/python-openstackclient/latest/cli/plugin-commands/gnocchi.html#metric-measures-aggregation","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":21692,"name":"licanwei","email":"li.canwei2@zte.com.cn","username":"licanwei"},"change_message_id":"bb1c16c009a29068c9bd42be44622470a104cbba","unresolved":false,"context_lines":[{"line_number":48,"context_line":"Proposed change"},{"line_number":49,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"To enable time series the datasource base class will need a new method. This"},{"line_number":52,"context_line":"method will allow metrics to be retrieved over a specific period with a"},{"line_number":53,"context_line":"specific granularity. This method will return a list of values according to the"},{"line_number":54,"context_line":"supplied parameters and will subsequently be implemented by all current"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_3ed91f22","line":51,"range":{"start_line":51,"start_character":0,"end_line":51,"end_character":70},"in_reply_to":"bf51134e_b8f57c59","updated":"2020-07-01 03:21:28.000000000","message":"so maybe you want to add new datasource?","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":21692,"name":"licanwei","email":"li.canwei2@zte.com.cn","username":"licanwei"},"change_message_id":"6119b9945a0bba7af15b09ce0dd3c45e172af0af","unresolved":false,"context_lines":[{"line_number":54,"context_line":"supplied parameters and will subsequently be implemented by all current"},{"line_number":55,"context_line":"datasources."},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"In addition, a new class will be added that can utilize a datasource to perform"},{"line_number":58,"context_line":"time series analysis. Features will include, decomposing trends, periodic"},{"line_number":59,"context_line":"variations (daily, weekly, seasonal) and irregular variations (residuals). As"},{"line_number":60,"context_line":"well as determining the stationarity. Any machine learning aspects such as"},{"line_number":61,"context_line":"predicting future values through models like ARIMA will **not** be part of this"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_78e00427","line":58,"range":{"start_line":57,"start_character":12,"end_line":58,"end_character":20},"updated":"2020-06-26 07:21:55.000000000","message":"Is it a common class used for all ML strategies?","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":29911,"name":"Dantali0n","email":"info@dantalion.nl","username":"Dantali0n"},"change_message_id":"119827acf2525135656d09ffb8780eb78401b317","unresolved":false,"context_lines":[{"line_number":54,"context_line":"supplied parameters and will subsequently be implemented by all current"},{"line_number":55,"context_line":"datasources."},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"In addition, a new class will be added that can utilize a datasource to perform"},{"line_number":58,"context_line":"time series analysis. Features will include, decomposing trends, periodic"},{"line_number":59,"context_line":"variations (daily, weekly, seasonal) and irregular variations (residuals). As"},{"line_number":60,"context_line":"well as determining the stationarity. Any machine learning aspects such as"},{"line_number":61,"context_line":"predicting future values through models like ARIMA will **not** be part of this"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_d8331013","line":58,"range":{"start_line":57,"start_character":12,"end_line":58,"end_character":20},"in_reply_to":"bf51134e_78e00427","updated":"2020-06-26 08:21:28.000000000","message":"No not many Machine Learning algorithms use time-series analysis. It is more commonly used in statistically methods such as simple linear or polynomial regression. Many of the algorithms used to support machine learning algorithms are actually from completely different fields such as signal processing or statistics. \n\nTypically these algorithms are only employed to extract features from the data or reduce noise but they should not be considered Machine Learning in itself. Which supporting algorithms are employed and which features are extracted entirely depends on the problem that is attempted to be solved.\n\nAn example were a time-series decomposition would be a great addition to the dataset used by a ML algorithm would be in weather prediction. As to predict the weather any seasonal change in temperature could essentially be noise in the prediction based on local deviations in the average temperature (residual).","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":29911,"name":"Dantali0n","email":"info@dantalion.nl","username":"Dantali0n"},"change_message_id":"eef82d6f400c24cd296a8259924585de8e5d7cac","unresolved":false,"context_lines":[{"line_number":54,"context_line":"supplied parameters and will subsequently be implemented by all current"},{"line_number":55,"context_line":"datasources."},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"In addition, a new class will be added that can utilize a datasource to perform"},{"line_number":58,"context_line":"time series analysis. Features will include, decomposing trends, periodic"},{"line_number":59,"context_line":"variations (daily, weekly, seasonal) and irregular variations (residuals). As"},{"line_number":60,"context_line":"well as determining the stationarity. Any machine learning aspects such as"},{"line_number":61,"context_line":"predicting future values through models like ARIMA will **not** be part of this"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_792c514a","line":58,"range":{"start_line":57,"start_character":12,"end_line":58,"end_character":20},"in_reply_to":"bf51134e_7e1277b6","updated":"2020-07-01 05:54:12.000000000","message":"Previously we discussed that actual ML strategies for Watcher would be best developed as external component.\nYet, I would still like to use time series decomposition for potential new strategies, these new strategies will not use ML only the time series decomposition. But, perhaps it is better to focus on developing these strategies in an external component so that we can keep the libraries for Watcher itself small.","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":21692,"name":"licanwei","email":"li.canwei2@zte.com.cn","username":"licanwei"},"change_message_id":"bb1c16c009a29068c9bd42be44622470a104cbba","unresolved":false,"context_lines":[{"line_number":54,"context_line":"supplied parameters and will subsequently be implemented by all current"},{"line_number":55,"context_line":"datasources."},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"In addition, a new class will be added that can utilize a datasource to perform"},{"line_number":58,"context_line":"time series analysis. Features will include, decomposing trends, periodic"},{"line_number":59,"context_line":"variations (daily, weekly, seasonal) and irregular variations (residuals). As"},{"line_number":60,"context_line":"well as determining the stationarity. Any machine learning aspects such as"},{"line_number":61,"context_line":"predicting future values through models like ARIMA will **not** be part of this"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_7e1277b6","line":58,"range":{"start_line":57,"start_character":12,"end_line":58,"end_character":20},"in_reply_to":"bf51134e_d8331013","updated":"2020-07-01 03:21:28.000000000","message":"From my understanding,  there are a lot of mature third-party software can use to process data and extract features, ML strategies can use these features directly, perhaps this is better because.","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":32161,"name":"Jatinder Singh","email":"jatinder1.singh@orange.com","username":"jatindersingh"},"change_message_id":"d1e8e2591d72695c64943d0d87eb0ef373d1ce84","unresolved":false,"context_lines":[{"line_number":116,"context_line":"------------------"},{"line_number":117,"context_line":""},{"line_number":118,"context_line":"Collecting lists of values instead of single values per node and or instance"},{"line_number":119,"context_line":"can significantly increase the memory consumption of Watcher. Especially, when"},{"line_number":120,"context_line":"audits are run for large collections of nodes and instances. Additionally, the"},{"line_number":121,"context_line":"transmission of all this data could potentially be a bottleneck if executed"},{"line_number":122,"context_line":"sequentially. For this the time series class could utilize the `decision-engine"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_63fa6975","line":119,"range":{"start_line":119,"start_character":18,"end_line":119,"end_character":59},"updated":"2020-07-09 12:30:12.000000000","message":"is there any consideration of environment resources utilization by bulk data metrics during collection/processing?","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"},{"author":{"_account_id":29911,"name":"Dantali0n","email":"info@dantalion.nl","username":"Dantali0n"},"change_message_id":"0945bad48b917b6213a2e0a6a5f44c1bcf3502c1","unresolved":false,"context_lines":[{"line_number":116,"context_line":"------------------"},{"line_number":117,"context_line":""},{"line_number":118,"context_line":"Collecting lists of values instead of single values per node and or instance"},{"line_number":119,"context_line":"can significantly increase the memory consumption of Watcher. Especially, when"},{"line_number":120,"context_line":"audits are run for large collections of nodes and instances. Additionally, the"},{"line_number":121,"context_line":"transmission of all this data could potentially be a bottleneck if executed"},{"line_number":122,"context_line":"sequentially. For this the time series class could utilize the `decision-engine"}],"source_content_type":"text/x-rst","patch_set":1,"id":"bf51134e_de27b111","line":119,"range":{"start_line":119,"start_character":18,"end_line":119,"end_character":59},"in_reply_to":"bf51134e_63fa6975","updated":"2020-07-10 14:38:30.000000000","message":"Primarily by audit templates, these allows to define a scope which limits the number of availability_zones and host_aggregates. Naturally this significantly reduces the number of nodes and instances (assuming compute datamodel) resulting in less memory consumption.","commit_id":"4e6896923a24431a399a6fdc0ce19fd1f92e02a3"}]}