Understanding "runtime mcycles" and "cpu_ms" accounting on AppEngine's Go runtime

I have a Go/AppEngine app that I'm trying to fine-tune to optimize concurrent requests, which is currently cpu-bound. In the process of doing so, I'm seeing what look like anomalous values for cpu_ms in the logs, and average runtime mcycles in the dashboard.

I have a few different endpoints whose cpu use seems completely at odds with reality, but one in particular stands out. It's a simple handler that reads roughly as follows:

	func ThangHandler(w http.ResponseWriter, r *http.Request) {
		ctx := appengine.NewContext(r)

		var orgId string
		cookie, err := r.Cookie(orgCookieKey)
		if err != nil || cookie.Value == "" {
			// Check URL params as a fallback.
			r.ParseForm()
			orgId = r.Form.Get("orgId")
			if orgId == "" {
				util.HttpError(ctx, w, http.StatusForbidden)
				return
			}
		} else {
			orgId = cookie.Value
		}

		w.Header().Set("Content-Type", "application/json; charset=utf-8")
		fmt.Fprintf(w, simpleTemplate, orgId, r.Host, "true", host)
	}

The details of this code don't matter so much as the fact that it doesn't do much more than read a cookie/param and run a Printf on a very simple template string (maybe 100 characters or so).

As I write this, the AppEngine dashboard is reporting this endpoint as consuming 83 runtime mcycles average over the last hour, which seems surprisingly high. When I look at the top 20 log entries associated with these requests, I see an even stranger picture. Most of them are either ms=13 cpu_ms=0 or ms=13 cpu_ms=21 (I assume there's some quantization going on there). But about 10% are really odd, such as ms=148 cpu_ms=238!

So my actual question is this:

• How can an endpoint this simple possibly consume 83 average mcycles, and have such high variance?
  • Should I suspect GC pauses?
• How is it ever possible that cpu_ms > ms in the logs?

For the benefit of anyone that comes across this question in the future, here is the answer given by dsymonds on the google-appengine-go mailing list:

> The cpu_ms and related accounting measures are legacy holdovers from
> the old billing structure, which was based at least partly on CPU
> consumption. Nowadays it is meaningless from that perspective, and I
> wouldn't be surprised if those numbers are somewhat nonsensical.
>
> There's nothing in the Go runtime done to attribute CPU time to
> separate requests, nor is it really tractable to do so in a concurrent
> runtime. The attributing is statistical in nature, which may account
> for the weirdness you are seeing.