$3 per dashboard per month (first 3 free) seems exorbitant. They're already charging for storing and viewing metrics, I would have expected dashboards to be free.

I actually hope they will utilize elasticsearch and dyanmoDB one day. They already have integration b/w DDB and Titan, DDB and Kinesis, Kinesis and Elasticsearch, so it's just a matter of time and enough requests to see Elasticsearch & Kibana integrate with cloudwatch. I am not sure when they will have a "logstash" although that's already doable with cloudwatch log.

But CW retention is still two-week maximum right? Would be really beneficial if AWS just make a service available or upgrade the two-week retention to customer-choosing. I know the metric storage can grow exponentially (like what happened with the recent Dynamo outage) but it will also alleviate customers to roll their own metric storage system if CW now can create custom dashboard.

Yes I want this too! You can pay them for CW log retention but not for metrics retention. Seems like an obvious step.

We just setup DataDog, and absolutely love them for all of our charting and monitoring needs. They are quite deeply integrated with AWS.

Do you have actually ship your AWS cloudwatch metrics into datadog? Or just events?

I'm curious if the costs of doing that (sucking all metrics from aws cloudwatch) start to get really high at scale?

We've been using DataDog for a while and I hate it. I'm really excited about doing this directly at AWS.

Can you elaborate on what's good / bad about it? We have been looking for a metrics solution and trying to find a good option.

What I'm confused about DataDog is they charge your per host - how exactly does that work if you are just reporting metrics (I guess you have to pay per host you are reporting metrics from).

Watching how players in the space are evolving how and what they bill for is interesting.

I think that charging by host is a lot easier for general buyers to understand and things generally average out.

I was recently also interested by the new signalfx pricing model - its really priced by metric volume but they sell it by host. When you peel back the covers they're just including a certain amount of metric volume with each host and your actual bill will vary.

Obviously underlying costs are tied to the amount of metrics, not the amount of hosts.

We are using Datadog for a bunch of stuff right now, and we generally like it. The per-host pricing model does get a bit hard to swallow especially when you have a bunch of really small hosts. Your monitoring bill can start to approach your hosting bill :)

Grafana (in v2.5) now has a bunch of overlap with this.

It's arguably a better solution [disclaimer: I'm biased; co-founder of raintank, the company behind Grafana].

Best of all its open source and used for metrics viz by a lot of people.

v2.5 ships with a new cloudwatch data source, so it's super easy to add cloudwatch metrics to your existing dashboards.

We'd love to get feedback on it.