Fork me on GitHub

I just purchased four new servers, and set out to find some good asset management software. Specifically, I’m looking for a PHP package that allows me to track the details of all of my hardware. While software and configuration tracking would surely be good for some applications, I already have a package to handle that.

I searched a few popular sites, including SourceForge, and came up with a dozen or so possibilities that looked good. They varied in level of detail and features, but all had one common failure - they lack an object-oriented or relational approach. What do I mean by that? Most of them are designed so that you enter data in a form, it goes into a database, and exists just as a table of data. Devices, networks, etc. exist independently.

What do I want? An object-oriented approach that can handle relationships between objects, and hierarchies. Lets design a few objects as examples: We’ll have Networks, Servers, Interfaces, Switches, Hard Drives, and UPSs. They each represent the real-world hardware. An interface, for our purposes, will be a physical way of connecting two devices - an NIC, serial line, etc.

If you’re experienced in object-oriented programming, you can think of relationships as inheritance, with an added capability of other references.

Our Server object will represent a real-world server. It has components (other objects) such as Interfaces, Hard Drives, and others.

How does this all pull together?

We define, in our asset management software, a Server which has one Interface (Ethernet NIC called “NIC1”) and three Hard Drives (called HDD1-HDD3). The Interface, in turn belongs to a Network (LAN/VLAN), and has a field called “connection” which references a Interface of a Switch object, specifically Interface “Port1” on Switch “Switch1”. We’ll define one Network called “LAN” which is identified by the 192.168.0 IP range (a field of the Network object). We create a Switch object, which belongs to LAN, has IP 192.168.0.2, and has 24 Interfaces, called “Port1” through “Port24”. This represents a BayStack 450-24T Ethernet switch. We’ll also define a UPS called “SmartUPS”, which in turn has an Interface which is part of “LAN”, Connected to Switch1 on Port21.

Now, you see the concept beginning to emerge. We have, in essence, an inter-related mesh of objects representing physical hardware and its’ properties. This can all be thought of in a three-dimensional form, which represents the connections and relationships in our network.

The main two advantages of this approach are:

  1. The ability to quickly recognize relationships between objects. If, for example, we looked at a summary page in Switch1, we would see a number of static fields describing its properties (manufacturer, model, IP, MAC, etc.) as well as a diagram of its’ relationships. Such a diagram would look something like:
Port1  -> NIC1 -> Server1
Port21 -> SmartSlotCard1 -> UPS1

It would also have a listing of Networks which this device is a member of, specifically:

LAN

In practice, on a web interface, each one of these entries would be a link to that object’s summary. Clicking on LAN, UPS1, Server1, etc. would show us the summary of that object, so that we can browse through our physical network. Clicking on LAN, for example, would show LAN’s properties, as well as all devices that are members of LAN (related to it).

The power of such an approach also relies on binding objects to specific pieces of hardware. For example, the disk drives in Server1 (HDD1, HDD2, HDD3) would in turn be references to objects representing actual physical assets. We can then move these assets around. For example, let’s say that we assign a unique serial number of 000306 at a physical hard drive. We now create an object for it, and reference HDD3 in Server1 to this object. If we reconfigure Server1 (physically) by removing HDD1 (000306) and place it on a shelf, we could then edit Server1’s HDD3 object to be “empty”, and have 000306 reassigned to an object “Shelf02” that represents a physical storage location. By viewing the summary for Server1, we would see that HDD3 is “empty”, but if we viewed Shelf02, we would see that 000306 is there, waiting to be used. When we add a server, Server2, we could assign 000306 to that server, as, say, HDD2.

To give another example, UPS1 is an uninterruptible power supply. We can “assign” devices to it, such as Switch1 and Server1. Viewing a summary for UPS1 would show us that Switch1 and Server1 are connected to it. Similarly, viewing the summary for either Server1 or Switch1 would show us that they are connected to UPS1. If we add UPS2, we could simply edit the Server1 object so that it is connected to UPS2, and the UPS1 and UPS2 objects would show this change automatically.

  1. Historical tracking. Every change would be accompanied by a person responsible, a date and time, a reason, and perhaps other information. Exactly how this would be implemented is not decided (whether relationships would be marked as ‘deleted’ or whether changes would be held in a separate table in the database), but the idea is that any object would also contain a history of relationships. From our previous examples, if we viewed a summary of Server1, we would see that the connection to UPS1 was severed on a specific date due to a reason such as “overload” and on that same date, the current connection to UPS2 was established. Similarly, if we view the summary for HDD2 on Server2, we would see that the object represents real-world hardware asset #000306. Viewing the summary for 000306 would show us that it is a 18.2 Gb SCSI drive in a Compaq hot-swap tray, and that it was originally installed in Server1, but moved on a certain date for a certain reason to Shelf02, and then later moved to Server2.

The implementation of this concept would most likely come about as a web-based PHP front-end with a MySQL database back-end. In order to achieve good functionality, there would be a simple interface with quick execution of common tasks. Most likely, the names I have used such as Server1, UPS1, HDD3 on Server1, etc. would just be reference names to a real-world hardware object, and its’ data object representation, identified by a unique ID. For example, the name HDD3 on Server1 would really just be a reference to 000306, a Hard Drive object.

If we physically removed this drive from Server1 and placed it on Shelf02, we would browse to the Server1 page in our web interface, and click on “HDD3”, which would bring us to a page representing 000306. We could then click on a “Move” link, bringing us to a form. This allows us to enter our name, date and time, reason, etc. and a new location of “Shelf02”. We click submit and the move information is entered into a table, which references all associated objects - Server1, HDD3 on Server1, 000306, Shelf02, etc. HDD3 on Server1 has the reference to 000306 removed, and a reference to a History object added. Shelf02 has a reference to the History object, as well as a reference to 000306, added.

Now, viewing HDD3 on Server1 would show that there is no physical device associated with it (empty), but would give a reference (link) to the 000306 drive, the History entry for the removal, and the current location of 000306 on Shelf02.


Comments

comments powered by Disqus