Integration Install Part 1

Well, I didn’t have time to write an original post yesterday, but I did post an interesting article on about Siebel Systems. At least I thought it was interesting, especially the part about them grooming themselves for a possible sale… especially since the name Microsoft was mentioned.

I was working on the Integration piece for my MSCRM and MS Great Plains for the last three days. I can’t wait to see the consulting bill on this one. It took a lot longer then they anticipated, and longer than it should have.

I provided them with the entire set of production dbs, well over 25 GB of data to test the integration with. I copied them over to an external HD and they came and picked it up back on Feb 22.

The installation architecture consisted of a SQL Server with a CRM db and a GP db on one box, and a SQL Server on another box that houses the Integrator, the idea is to have those on separate machines to help load balance. And even though this was the plan from the beginning the consultant only tested the integration on a single machine and a single SQL Instance. Needless to say there was some inconsistencies, and it had to be redone a couple of times.

Testing and evaluating are important factors on any install. One of the key points to evaluating a system is to ensure that as much of the system as possible is reproduced on the test server or in the test environment if more than one server is being tested. The ultimate would be to test it on the exact same hardware, but for the most part that’s not feasible. If you plan on running a piece of software on a server that runs enterprise Antivirus, you better be sure to put it on there when you test. If a consultant or VAR is testing something for you be sure to push them to get the environment as close as you can. Its your money and testing can reveal a lot of problems and lets you try things several times before paying three people to sit and troubleshoot an install issue that should have been identified in testing.

Anyway, our last integration piece was the default BizTalk model that was released 6 months after CRM came out of beta. And like I’ve said here before, it worked – but just barely. It was slow, and required more resources than necessary. And when large updates were pushed through like the company wide price increase we had last year it broke in the middle and had to be left behind. Literally. It took us weeks to fix it, upgrade to 3 GB of RAM, then another 2 weeks to force feed the info through the sausage grinder. We had to break it down into bites and feed it only at night. Feeding it during the day slowed the production servers that ran GP down so much that we couldn’t post invoices or run our BOM inventory auto builds.

The initial integration took about 15 hours with that monstrosity. Once the tedious installation took place, Scribe Insight shattered that time, even though the database had grown considerably since the initial go around 2 years ago. I’m not exactly sure how long it took, because It was done the first time I checked on its progress 5 hours later. In fact it was done so quickly I initially thought it had failed until I saw that it had update or integrated 25,000 records in the monitor.

I’ve noticed some periodical slow downs, and some minor problems on a few records, but for the most part I feel pretty good about the integration. The true testing starts tomorrow when the consultants come back to look through the results, and train me a little further on the GUI. I’ll keep you posted on the results.