Skip to content
Permalink
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
118 lines (72 sloc) 3.78 KB

Algorithm Graphs Analytics

This section will expose how analytics graphs works.

Analytics system follow the steps below.

Maestro Server - Analytics maestro architecture


Making mongodb graph lookup

Graph lookup will create a python dict using mongodb graph, and all application entities have a dep field, its a list with app objects [id, name].

../_static/analytics_graphs/ana_mongo.jpg


Creating a networkX graph

After graph sends to network business, basically transform python object to networkX.

We have a recursive function inside each node on the tree, this order will apply a lot of rules creating a new graph tree with better UI order, like in these moments we handle with conflicts, double dependencies, orders and etc.

Maestro Server - Analytics Recursive

Code example showing the recursive function

def _recursive_draw(self, app, i=0, OHelper=HelperOrderedSuccers):
    if i > 30:
        return

    for item in app:
        if not self._grid.in_index(item):
            node = self._graph.nodes[item]
            helper = self.add_pos_grid(node)

            succ = OHelper(helper).get_succers()
            self._recursive_draw(succ, i + 1)

In these graphs we can see all rules applied on graph to create a best dependency tree. Each recursive tick try to figure out certainty pattern, if found apply a transformation. For example The child balance pattern try to find a multiple child in one parent, if happen then create a dummy item on parent layer.


Rules

Growing node

  • When: If the node have more than one child, growing the node to be equal of the number of child
  • Transform: Set the node size to be equal of number of child

Child Balance

  • When: If the parent node have more than two child.
  • Transform: Create a dummy item beside to node parent.

Chess Pawn

  • When: If the app it's an entry point and have parent.
  • Transform: Skipped one column

Maestro Server - Analytics

Chess horse

  • When: If the node have a top obstacle which other nodes point out to a common dependency.
  • Transform: First push back the dependency to a clear column, and after create a dummy path to the new column.

Maestro Server - Analytics chess rules

Clear rows

  • When: After finish the recursive loop, figure out empty columns.
  • Transform: Delete these column and rebalance the grid.

Maestro Server - Analytics clear system


Enrichment data

Next step is an enrichment layer with getting all server data used by each graph.

The enrichment step received two dataset a json python dict represents a graph tree, a matrix position grid. One worker will get all applications accordingly with graph tree creating a new dataset called servers.

Maestro Server - Analytics Enrichment


Draw layers

Now its draw time, where we have three datasets, one its a graph tree, second its matrix od each position and the last its list of servers.

Maestro Server - Analytics Rotation

Maestro Server - Analytics Vertical

You can’t perform that action at this time.