Summary
<%
var hits = blitz.pp.percent(test.last_point.hits*100/test.last_point.total);
var errs = blitz.pp.percent(test.last_point.errors*100/test.last_point.total);
var tuts = blitz.pp.percent(test.last_point.timeouts*100/test.last_point.total);
var hitsp = hits !== '0' ? hits + '%' : '';
var errsp = errs !== '0' ? errs + '%' : '';
var tutsp = tuts !== '0' ? tuts + '%' : '';
%>
This
rush
generated
<%= blitz.pp.number(test.last_point.hits) %>
successful hits in
<%= blitz.pp.duration(test.max_duration) %>
and we transferred
<%= blitz.pp.bytes(test.last_point.rxBytes+test.last_point.txBytes) %>
of data in and out of your app.
The average hit rate of
<%= blitz.pp.rate(test.average_hit_rate) %>
translates to about
<%= blitz.pp.number(Math.floor(test.average_hit_rate*3600*24)) %>
hits/day.
<% if (test.average_response_time > 0.5) { %>
The average response time of
<%= blitz.pp.duration(test.average_response_time) %>
is considerably higher than most other sites that are built
to scale out. Response times less than
250 ms
are what the cool kids strive for.
<% } %>
<% if (test.last_point.timeouts || test.last_point.errors) { %>
You got bigger problems though:
<%= ((test.last_point.timeouts+test.last_point.errors)/test.last_point.total*100).toFixed(2) %>%
of the users during this
rush
experienced timeouts or errors!
<% } %>
<% if (test.first_error_point) { %>
Errors
The first error happened at
<%= blitz.pp.duration(test.first_error_point.timestamp) %>
into the test when the number of concurrent users was at
<%= test.first_error_point.volume %>.
Errors are usually caused by resource exhaustion issues, like running
out of file descriptors or the connection pool size being too small
(for SQL databases).
<% } %>
<% if (test.first_timeout_point) { %>
Timeouts
The first timeout happened at
<%= blitz.pp.duration(test.first_timeout_point.timestamp) %>
into the test when the number of concurrent users was at
<%= test.first_timeout_point.volume %>.
Looks like you've been rushing with a timeout of
<%= blitz.pp.duration((test.timeout||1000)/1000.0) %>.
Timeouts tend to increase with concurrency if you have lock contention
of sorts. You might want to think about in-memory caching using
redis,
memcached or
varnish
to return stale data for a period of time and asynchronously refresh
this data.
<% } %>