You need a sandbox

To become good at building reliable solutions you have to be really good a breaking things. You must have a sandbox that allows you to play, experiment, and develop an intuitive sense of how things work. Over the years, I have literally spent thousands of hours in sandbox environments making a mess of things so that I can learn how not to make a mess of things. For me, there isn’t any substitute. You can’t just read it in a book, take a class, or work solely on production systems. You must have a place where you can break things and take chances.

Here are my recommendations for building a first class development/learning environment. From the PI Geek perspective, these recommendation will help for someone who is interested in developing solutions, or for someone who is interested in learning to deploy these solutions. The faster you can try things and recover when you mess things up, the faster you can learn.

MSDN is a must

First off, an MSDN subscription is a must. You are going to need a number of licensed Microsoft components in order to configure a complete development environment. Off the top of my head, I can say that you need Windows Server, Microsoft SQL Server, SharePoint Enterprise, and Visual Studio Professional. You need to be able to get your hands on these things whenever you need them. The best way to get all of these is to have an MSDN Professional subscription. I have been able to get this subscription for about $1200/year by committing to a 3 year deal. If you want details, just shoot me a message @RaesemannEnt and I’ll point you in the right direction. If you know of a better deal, please tweet me. I’ll promise to buy you a beer with some of the savings.

Go Virtual or Go Home

Next, go virtual or go home. Adopting virtualization is probably the single largest productivity increase that you can give yourself. The key to this productivity boost is the fact that you can easily back up your environment prior to making changes by taking what is known as a snapshot. This snapshot preserves the state of the system at that point. You can keep multiple snapshots making it easy to restore to different states.

Any of us who have played video games over the years has already learned this snapshot concept. You save your game progress at a dangerous point in the game so that you can start over from that point when you die. Notice that I didn’t say if you die, you always die the first several times you try. This concept of a save is not only convenient, it becomes a key strategy in learning to play the game better. Saving the game allows us to test different strategies and tactics at very little cost. Nobody wants to restart from the very beginning of the game when they die. Games that don’t let you save just end up pissing you off, so don’t play that game. Virtualization gives you this edge in real life.

I’ve used VMWare Workstation and Server for many years. I have recently, in the past year or so, decided that Hyper-V is a better solution. Hyper-V is the virtualization technology that is built into Windows Server. In my case, I use Windows Server 2008 R2 which allows you to add Hyper-V as a server role. Once you install that role, you are ready to roll. I prefer Hyper-V primarily because I find that it uses the host system resources more effectively than VMWare. In other words, I can run more virtual machines and see better performance under Hyper-V than I did under VMWare server. You can see from the screenshot below that you can stack quite a bit on a decent box and still get acceptable performance.

If you are going to be a good PI Geek you have to live on the edge. That means that you are going to download the newest available software and be the first to play with it. Betas can be ugly. I have found that while the end OSIsoft product is usually in really good shape by the time they release it to beta, the installer/installation process can have all sorts of complications that don’t pop up until it has been installed in a lot of different environments. Since you are the beta tester, that means that you are likely to be the one that discovers one of the problems. You are going to be glad that you made that snapshot prior to testing the betas.

In part 2, I’ll discuss some other strategies for installing different OSIsoft products. I am growing especially fond of Windows 2008 R2 Core as the basis for my virtual machines and I am slowing starting to migrate things over to that. You can read about my experience installing the PI Server 2012 beta in core here.

This article in SharePoint Magazine covers the CAML queries but I found that examples given produced an error

"Cannot convert System.Xml.XmlNode to System.Xml.Linq.XElement"

In this example I was trying to query a list for all list items that did not have a “Status” of “Completed”. Here is the code that finally worked.

var spquery = new System.Xml.Linq.XElement("Query",
 new XElement("Where",
 new XElement("Neq",
 new XElement("FieldRef",
 new XAttribute("Name","Status")),
 new XElement("Value"
 new XAttribute("Type","Choice"),"Completed"))));

var viewFields = new System.Xml.Linq.XElement("ViewFields");
var queryOptions = new System.Xml.Linq.XElement("QueryOptions");

XElement items = ws.GetListItems("Example List",
 "{D4200E79-28AF-4DF0-A106-499226DD38DB}",
 spquery,
 viewFields,
 null,
 queryOptions,
 null);

This produces the following XML for the query…

<Query>
 <Where>
 <Neq>
 <FieldRef Name="Status'" />
 <Value Type="Choice">Completed</Value>
 </Neq>
 </Where>
</Query>

I downloaded the PI Server 2012 Release Candidate yesterday and decided to setup a test VM so that I could play with it. I’m working on a new blog post about my vCampus development environment where I describe how I’ve chosen to set it up in more detail. For now, just know that I am running everything on a Windows 2008 R2 Server with Hyper-V enabled. I am moving all of my VM’s to Windows Server Core because it has a smaller footprint allowing me to run more VM’s on my host with better performance. It also has the added benefit of needing fewer Windows Updates making the entire environment easier to maintain. As an added bonus, I am getting a lot of experience with server core and PowerShell scripting.

It turns out to be relatively trivial to install PI Server on Windows Server Core. First you need to enable the .NET features on the server. It is possible that you don’t need all three of these but it only takes a few minutes to do it and this is just a development environment. For a production system, I would probably confirm that they are all needed and leave out any that are not absolutely necessary. Here are the command lines to install these features.

dism /online /enable-feature /featurename:NetFx2-ServerCore-WOW64
dism /online /enable-feature /featurename:NetFx3-ServerCore
dism /online /enable-feature /featurename:NetFx3-ServerCore-WOW64

Once these features are installed on the server, you need to run the OSIsoft Prerequisites installer for your platform. You can then run the PI Server setup and follow the directions. The GUI installer works exactly like a regular installation.

I found that you can even run the SMT gui from the server as you can see below.

I’m going to try to set aside some time to try some of the new features of the server. I am especially interested in how much easier it is going to make backfilling old data into the server.

The presenters were very clear that much of what was discussed in this presentation was preliminary work and said that it was nice to be talking about something without having to discuss anticipated ship dates. Some of this is in in early proof of concept stage and some is closer to market.

OSIsoft is planning to leverage the Microsoft Windows Azure platform as the basis for their cloud offerings. This makes perfect sense because this means that they are able to port their existing code base. All of their experience developing for the .NET platform using Visual Studio is applicable. I’m not sure how much work is involved in porting to Azure but I suspect that time to market for some of these products may be fairly fast as they are only making modifications to existing code rather than developing a large new code base.

I haven’t personally had any hands on experience with Azure. My MSDN subscription gives me some free cloud resources to use, but I haven’t had any business motivation for devoting any time to it yet. I am under the impression that Azure really encompasses a wide range of offerings, from allowing you to host some data and web code on the cloud on up to running Hyper V VM’s of your systems.

There are really two main problems that the cloud might help us with. The first problem is the need to distribute PI data from one source to various destinations. The second is the need to distribute PI data to the rapidly growing number of smart phones, tablets, and consumer type devices. In both cases there are existing solutions, but the cloud potentially offers some advantages, such as the ability to provision these services quickly on an as needed basis, without the need to purchase new hardware, or reconfigure existing networks.

Many customers have had the need to transfer PI data from one site to another for quite some time. There are any number of reasons for this. Examples would be sending data from the source up to the corporate office, sending data to a business partner, or sharing data with a customer. Today this usually means setting up VPN tunnels and a server to host the PI to PI interface. OSIsoft is working toward being able to offer a range of options, from hosting the PI to PI transfer in the cloud, to hosting PI servers in the cloud. It isn’t entirely clear whether they will just provide the tools for you to set this up yourself, or if they will have some sort of service offering.