- Tools need to show updates in seconds (like, 10).
- Feature Flags are a term of art.
- monitoring needs to be more available and scaleable than the thing being monitored
- measuring business value
- Problem detection and analysis
- "Ultimately business value is what the business values and that is that"
- Possible values:
- Customer happiness
- Cost efficiency
- safety and security and compliance
- Customer happiness: Time to value, availability, response time.
- Efficiency: High utilization, Optimized, automated
- Why isn't this solved?
- everything keeps changing (interfaces, metrics, schema, scale, ephemerality)
- and serverless...
- Cost per node drops, revenue opportunity decreases, continuous disruption. Each generation drops the price by an order of magnitude
- Tragic quadrant.
- Simulator of large scale architectures. spigo github.com/adrianco/spigo
- Measuring response times in spigo (and the distribution of the same)
- Guesstimate http://wwwgetguesstimate.com/models/1307
- Lambda, Cloud Functions, Azure Functions (latter are beta)
- Monitorless?
- Good for simple, single threaded code.
- Security model is whitelisting.
- Doesn't live long enough to be exploited... Doesn't live long enough to be monitored (classically).
- Low traffic apps and super spiky apps are huge wins for serverless
- Tool for ease of use
- Must do multi-region
- Debugging and testing are... interesting
- Monitoring (end to end tracing?)
- Using Lambda to monitor AWS
- IBM is selling a product for this... for real.
- Scheduling and startup latency needs tackling
- execution and monitoring overhead
- how do you charge for this?
- capacity planning
- Too much new stuff
- Too ephemeral (and getting more so)
- price disruption (Whats the cost model to monitor a lambda function)