Geowave is a tool to store, index and search Geographical data on top of the key-value pair data storage. Geowave is built on top of the Geoserver and work on top of the apache accumulo. Apache accumulo is one dimensional data representation using key value pair. Geo wave provide the way to Multi Dimensional data representation to accumulo from Geowave. Also with that it get the capabilities to do the Geo Graphical data operations to the Accumulo data. Apache Accumulo provide the Accumulo overview page to provide Statistical data based on the Accumulo Data. It include the Table, Tablet and Tablet server based statistics. In this project, idea is to provide same set of statistical information in a different manner. Using same Data, represent them as a Geospatial information will give and very descriptive and easy interpretation to the users. Thus, using Geo spatial information represent Stat related to the Tablet and Tablet information display on a Map.
Solution to this scenario was implementing Map-based Monitor for Apache Accumulo. Application is run as web application and implemented as extension to the Geowave. Which connect to the Accumulo and related component to extract and fetch Stat related information using Existing or implemented REST api. Fetch data is again reformatted and interpreted into Geographic information to display it on a map. Which should contain the Table and Tablet related Statistical data and vice verse. Application implemented as a two component containing web application and REST api. Web application use the implemented or existing REST api to fetch and reformat the data. Both of the component is integrated with the Geowave at the end.
Architectural solution for the above solution is described as below.
this component is responsible for extracting statistical information associated with tables. Here it use Accumulo inbuilt implementation from “MasterMonitorInfo”. Using it’s “getMasterStats()” method it is possible to extract the accumulo related statistical information. It contain the most of the Statistic information including tablet,table and tablet server stats. To filter the table related statistic, here it use the “tableMap” property. It contain the all the tables related stat data. Next data again moved to the Jax-bean called “TableBean.java”. Data is set into the “TableBean” after the data parsed to the appropriate data formats.
This component is responsible to extract the Tablet server related statistic data. Tablet servers are physical nodes which are storing data chunks(Tablet) of the tables. For this as well it use the “MasterMonitorInfo” class. To filter the Tablet server related data it use the “getTServerInfo()” method to filter it out. It contain the list of “TabletServerStatus” component by holding table statistic related to each tablet server. Using the above method it is not possible to get the tablet server statistic data directly. Thus what is has to do is, it need to aggregate all the values of the each table in the each tablet servers to get the final tablet server related statistic data. After the data are extracted it is formatted and moved to “TabletServerBean” object.
Tablet Stats component is extract the Tablet related statistic. Tablets are physical representation of the Logical tables. Logical table is represented as collection of the Tablet by dividing the table row wise. These tablets are assigned to different tablet server randomly. Statistic extraction implementation done using both “MasterMonitorInfo” class and “TabletClientService.Client”. These data again formated and represented as “TabletBean” class.
Namespace operation is responsible to retrieve the namespace related table information. Namespace is something specific to the Geowave. It is a standard Accumulo Table. But Geowave create specific Table which data representation is different from Accumulo data. Accumulo initial data representation is still preserved here. But Geowave define additional parameters and column family to support Geowave operations. In that case, to identify the Geowave related table here they use namespace. Namespace is identified by the string value. it is append to the standard table name when create it. if the table going to create us “SPATIAL_VECTOR_IDX” and if the namespace is “namespace”, the result table name would be “namespace_SPATIAL_VECTOR_IDX”. In this module extract the namespace from a given table. For that it use the above strategy use to compose the name. Other function is return the list of tables corresponding to the given namespace.
This module is the most of the important module in the application. Because which is the one who do the most important task of the project. What is does is finding the geospatial extent or Geospatial representation of the Tablet and formatting again it to show in map. It there main functions. Those are, finding the table range, getting geospatial extent based on range, calculate convex hull of geospatial extents.
First it find the all the split points of a table. Using that it find out the table of the table. Next using the tablet it find the key range of these each tablets. Next use this ranges to find the geospatial extent of the tablet. Then iterate all the rows of this tablet to find the geospatial extent using the “AccumuloUtils.decodeRow()” function. Geospatial extent contain the list of coordinates representing the corresponding tablet on a map. Next using these points find the convex hull of the tablet. Convex hull is calculated using the “com.vividsolutions.jts.algorithm.ConvexHull;” class and associated functions.
function of this module is to start background daemon thread to find the spatial extent and convex hull of each of the tablets of the tables. Purpose of making them as background thread is above mention calculation are time consuming functions. Because of this it calculate the convex hull periodically and hold it to server when new request instantly.
First it find out the all the tables of the Accumulo and which support Geowave operations. TheAl then It find the all the tablet convex Hull and move to the List. At the end this module contains the map holding above calculated list of convex Hull. Map keys are corresponding to the table name and values are corresponding to the list of tablet convex hull of that corresponding table.
This calculation is start up when wep application loaded in to the container. This module implemented class is “ServletContextListener” class and “Thread” class. It’s “contextInitialized()” start to execute when it loaded it to container. This is the place it start to execute calculations. Above function is overridden and have implemented the above mentioned calculation. Also it contain the function “contextDestroyed()” which fire up when application destro or unloaded. ALl the execution interruption function are call inside this. In “contextInitialized()” method has defined the scheduler to define the time period to perform convex hull calculation.
Above mentioned function are exposed as rest service. Then this REST service or API can be used to any other extending features. Web application is also implemented using this REST API. All the API is implemented and server as HTTP GET methods.
Here are REST api information in brief.
which is responsible to display Table and tablet statistic data. It is possible to see the Tablet information filtering according to the any namespaces. when click the one of the table name is shows the all the tablet extent on the map. it is possible to hide and show each of this table on the map. And when mouse hover on each tablet, it will popup with the tablet statistics.
Here all the interaction are done via the REST api. This web application is completely running at the browser. when it need any statistic or geospatial extent information, it make REST call to REST module to fetch them and display on the client web app.
Above mentioned REST module and web application are work together to implement Geowave monitor. REST module which is java web-app implemented using jersey, is depoy inside the geowave web-app module. Also UI interface is put inside the webapp module. But it can be disassemble any time and host on anywhere else. Because it is not tightly couple with Geowave and it just use the REST api. After the build the geowave web-app module is it possible to deploy on the java web app container and run.