Search code examples
databasedatabase-designgraph-databases

Is Graph database really good for monitoring IT enterprise?

I have a simple question about graph database utility to monitor IT enterprise.

Many graph database vendors are suggesting that graph databases are optimal for monitoring IT resources. Such like Ref 1 or Ref 2

In order to monitor IT resources, it is necessary to manage log data from IT enterprise effectively.

However, there are many books for graph database modeling that leads us to model by the following time tree method.

enter image description here

I think that the time tree method will generate too much edges and will not be good for the performance of the graph database.

So... Is Graph database really good for monitoring IT enterprise ?

Solution

  • In my opinion, it seems that the questioner is confused with "the monitor the IT asset as an example of the excellent aspects to trace the complex relationship" and "the monitor log data for device history management".

    For an IT enterprise that provides numerous services, if one logical or physical system fails, its impact can affect other assets associated with that asset. (This is similar to the disease spreading in networks.)

    The graph database is very appropriate for configuring the system to track it. (please find the following figure) enter image description here

    Next, let's talk about the writer's question, it is a little more fundamental, in order to identify the cause of a specific malfunction in IT asset management, analysis of log data of asset should also be done. In this part, I understand how to model time series data on a graph database.

    In conclusion, RDB is better than GDB in the area of log data analysis for history management. (To be specific, Time series database is the best.) The reason is also for the same reason as the questioner mentioned. (It is true that Graphdb is better than rdb in terms of edge travers, but paradoxically, modeling to generate too much edge to this blind faith that is rather exerts an adverse effect on performance. This is called Dense network problem.)

    Therefore, it is recommended to implement the log data analysis function by constructing it in time series database and linking it by using DB Link, FDW, or using multimodel db of GDB + RDB such as AgensGraph, Oracle or OrientDB.