So You’ve Been Asked To Manage AI in GCP…
No big deal - just add it to what you’re already doing…
In the past year and a half of Customer conversations there has been an increasing trend of Executive Suites mandating AI adoption. How? Delegated. Iterative improvement? That’s what the AI is for! Kidding aside - it’s a real thing, and teams need to figure out how they’re going to take on management of new types of workloads.
Hyperscaler sellers are selling directly into business units, so the same is true for when IT leaders learn that any given department is proceeding with adopting AI without going through IT first.
With that in mind, today we’re deviating from desktops amd starting a new series covering how existing IT leaders can take on management of AI initiatives. This can be done in a way that’s relatively familiar to what an IT leader’s teams are already doing - while coming out looking like a LEGEND to your Executive Suite to boot.
Google’s tools with Vertex AI have proven impressive and thoughtfully built. YES, the concept is new. YES, it’s different. NO, you probably shouldn’t build it yourself if you’ve never done it before… but YES, you CAN monitor it and NO, you shouldn’t immediately bring your existing tools along for the ride.
While 3rd party licensing like Splunk (logging what AI is doing is important!) might be part of your stack today, I’ve historically advocated that you use native PaaS services instead for the purpose of easy integration.
No, not that kind of monitoring!
IT teams and/or MSPs have been monitoring systems forever. With the advent of RMM tools, laptops and servers could be monitored, patched and rebooted from anywhere. Troubleshooting became a breeze, because you had so much data at your fingertips - without driving across town.
I’m going to break down how Google Cloud Platform tools allow your team to recreate that approach - you’ll be able to have that same Tier II Server Admin monitor yet another workload without panicking, stalling or going through an infinite Learning course blitz (that doesn’t match the hands-on, IRL experience anyway).
Key points that we’re going to replicate…
Performance tracking & analytics
Log aggregation
Alerting
The cohesive management experience
Google Equivalents
Vertex AI = your new “RMM”. We won’t use this traditionally - instead of CPU and RAM, we’re monitoring TPU and things like latency. This also includes your alerting mechanisms. As long as you know what your thresholds are and who you need to notify, you don’t necesarily need to understand everything about the underlying model.
Google Cloud Security Command Center (SCC for short) = SIEM tooling - logging, threat and anomaly detection. This is often going to replace something like Splunk, meaning you’re going to avoid costs from adding another workload to an already expensive toolset.
Chronicle = analytics, including AI-powered log analysis
In Summary…
Vertex AI works a LOT like the RMM tools you’re used to in RMM tools, displaying critical performance information.
Leveraging the native PaaS ecosystem means everything is integrated within GCP, and that means speed. When AI is making decisions in real time, getting insights quickly matters A LOT.
Usage-based pricing for SCC/Chronicle can deliver notable savings over traditional Splunk pricing
Business as usual…
In summary, don’t be afraid to take on management of this new type of workload - the concepts you’ll need are natural extensions of what your team is already doing! IT leaders and their team can get hands-on experience and the feather in their cap for taking on AI initiatives. There’s no substitute for practical application, after all!
So, now you’re managing and monitoring Vertex AI in GCP with Cloud Operations Suite! The dashboards are as familiar as your old RMM, all with a ‘1+1=3’ edge from native PaaS integrations.
Were you curious about what performance metrics we’re tracking, if not CPU and RAM? That’s a post for another day!