It’s time to face facts; PowerCenter is now so good that trying to make a mapping go faster by performance tuning the mapping alone isn’t worth it.
OK, if a mapping is poorly written to begin with then it needs to be fixed but that’s not performance tuning: that’s correcting what should have been right first time around. So since Big Data is bound to mean bigger performance challenges, just what DO you do when all of the obvious gotchas are eliminated and the mapping is still not fast enough?
That’s the issue we faced recently. We had squeezed every last drop of inefficiency out of the code but it still wasn’t enough to meet the SLAs on the biggest CRM implementation, at the time, in Europe. We had to look outside Informatica but where should we begin? Pointing the finger at Network or Database or Storage performance is an unpredictable and dangerous sport, never more so when a huge amount of money has already been spent on the infrastructure, as it had here. And anyway, who’s to say it’s any one thing: it could be a combination of factors? And the infrastructure teams often start from the principle that performance issues are the application’s fault, right? The very least they will do is to question our competence to question them. But there was nowhere else to go: we needed to find a way to present evidence, not opinions, that performance might be improved. Otherwise it was goodbye to our SLAs.
We needed something that spoke ‘both languages’: a trusted performance monitoring tool that could present results based on proven knowledge of every part of the infrastructure and how they interacted with PowerCenter when we ran our schedule. But was there a tool out there that fitted the bill? Yes, if you wanted generic Infrastructure information but no if you wanted to view this information in the context of PowerCenter. There was nothing out there could tell us, for example, exactly which mapping of the four running simultaneously at 2 am was consuming most CPU, how were their parent workflows being distributed across the grid and was network or storage latency varying over time and which workflows suffered most? We needed to be able to speak to the Network, Database, Server and Storage Administrators using statistics that were generated by those parts of the infrastructure that they know and trust. We knew that it wasn’t a dense report we needed but easily communicable graphs of infrastructure performance in the context of the PowerCenter workflows, sessions and mappings that we were running. And we needed them in near real time, not two weeks later.
Does this sound familiar to you? Don’t you agree that it is time to rethink what Informatica Performance Tuning should actually mean? PowerCenter is mature enough to enable us to design and write mappings for optimal performance first time around making performance a given, not something we retrofit. That done, if you have performance challenges then you can start by looking outside of PowerCenter and engaging the other people involved in meeting the challenge: your infrastructure colleagues.
We did just that and we never looked back; we even became a trusted partner of Informatica along the way. So what changed? Well, we never found that elusive tool, it didn’t exist, so we wrote our own and called it HIPPO. Now we could share a view of Informatica performance that the infrastructure folks recognized and accepted. They could help us meet our SLAs because they could literally see our problem. The steps we took from there backed up our conviction that real progress in tuning means finding a common language to share an understanding with infrastructure teams of how PowerCenter interacts with the infrastructure that it sits within. Finding a common viewpoint unlocks their insight and expertise and helps everyone understand the real problem areas.
Big data means bigger performance challenges for everyone across IT’s great divide so if you want to be ready to meet these challenges then get a fresh perspective by test-driving HIPPO. You’ll be amazed what can happen when everyone can see the big picture.