In my previous posting on The Fallacy of Minimum Hardware Requirements I talked about the differing motivations and expectations behind a set of Minimum Hardware Requirements for an application and I ended the thoughts by saying that one should question any such set of requirements that is greater than 3 years old. In November 2004 Open Text published a series of White Papers called "Livelink Minimum Hardware Requirements for 100- or 1000-User Populations" for the four main infrastructures of the day (Win2k with SQL, Win2k with Oracle, Solaris with Oracle and HPUX with Oracle) and although I was running the team that created these papers and I know that they remain relevant in some senses, I think the wisdom they try to provide should be questioned. Now questioned doesn't necessarily mean it is wrong, it just shouldn't be taken at it's face ... a lot has happened since 2004.

The first thing we should do is recognize a few facts:

• in 2004 "Livelink" was the name of the core product sold by Open Text ... there were a few other products back then, even the Archive Server was part of the family, but without a doubt, the lion's share of the revenue came from the application called Livelink. So, while these papers may yet have value we do know that they are not sufficient for they don't speak to the Archive Server let alone the many changes to the application itself.
• there are no version numbers associated with the "Livelink" mentioned in these papers...that was a deliberate action, the intent was to create a reusable document that would be suitable for multiple versions
• the documents speak to specific scenarios that were felt to be representative of Livelink's use at the time but very much contained Livelink to a Document Management system

The second thing to notice is what is done correctly within these documents:

• a model ("TIPS" or Transactions Initiated Per Second) is presented as an abstract concept that provides an explanation of how and why this set of requirements were chosen (in other words, it is clear that these numbers were derived and not simply the best guess of experienced consultants)
• a model (Usage Profile) is presented as an abstract concept that provides the ability to gage how applicable these results are to a specific installation
• two distinct but specific scenarios are provided that are an order of magnitude apart in size (100 and 1000 users) allowing readers to choose something relevant for their solution
• the testing environment is provided (and is the clearest indication of just how dated these papers are)

The third thing to focus on is how this document contains outdated assumptions or missing information:

• in 2004 a 5-second response time may have been acceptable for authenticated applications like Livelink, but in 2010 5 seconds is a lifetime
• the base Library-centric Usage Profile may well still be valid for actual customer use of Livelink (though I find that unlikely) but does not represent any of the current applications and solutions currently being sold by Open Text
• while the number of TIPS was a defensible function (1% of named user) there was almost no mention of the expected CPU usage/requirements or timings short of the aforementioned and questionable "within 5-second response time" per transaction

Finally, the conclusions / requirements provided are ridiculous on their face as they talk about hardware specifications that are simply unavailable for purchase anymore and even used equipment using Pentium III specifications would be difficult to come by at best.

There is much about these White Papers that remains valid and useful but there is much that must be ignored. Unfortunately, if there are more recent versions of these documents they are very difficult to find. So effectively there is no Minimum Hardware Requirements available and thus a different set of metrics are being used and abused to form designs.

So, what are the minimum hardware requirements for Livelink then? I'll explore that question in a separate posting.