README.md in pupa-0.0.7 vs README.md in pupa-0.0.8

@@ -43,9 +43,59 @@
 
 1.  If you scrape many types of data from the same source, you may want to split the scraping into separate tasks according to the type of data being scraped. See [bill.rb](http://opennorth.github.io/pupa-ruby/docs/bill.html) for an example.
 
 1.  You may want more control over the method used to perform a scraping task. For example, a legislature may publish legislators before 1997 in one format and legislators after 1997 in another format. In this case, you may want to select the method used to scrape legislators according to the year. See [legislator.rb](http://opennorth.github.io/pupa-ruby/docs/legislator.html).
 
+## Performance
+
+Pupa.rb offers several ways to significantly improve performance.
+
+In an example case, reducing file I/O and skipping validation as described below reduced the time to scrape 10,000 documents from 100 cached HTTP responses from 100 seconds down to 5 seconds. Like fast tests, fast scrapers make development smoother.
+
+The `import` action's performance (when using a dependency graph) is currently limited by MongoDB.
+
+### Caching HTTP requests
+
+HTTP requests consume the most time. To avoid repeat HTTP requests while developing a scraper, cache all HTTP responses. Pupa.rb will by default use a `web_cache` directory in the same directory as your script. You can change the directory by setting the `--cache_dir` switch on the command line, for example:
+
+    ruby cat.rb --cache_dir my_cache_dir
+
+### Reducing file I/O
+
+After HTTP requests, file I/O is the slowest operation. Two types of files are written to disk: HTTP responses are written to the cache directory, and JSON documents are written to the output directory. Writing to memory is much faster than writing to disk. You may store HTTP responses in [Memcached](http://memcached.org/) like so:
+
+    ruby cat.rb --cache_dir memcached://localhost:11211
+
+And you may store JSON documents in [Redis](http://redis.io/) like so:
+
+    ruby cat.rb --output_dir redis://localhost:6379/0
+
+Note that Pupa.rb flushes the JSON documents before scraping. If you use Redis, **DO NOT** share a Redis database with Pupa.rb and other applications. You can select a different database than the default `0` for use with Pupa.rb by passing an argument like `redis://localhost:6379/1`, where `1` is the Redis database number.
+
+### Skipping validation
+
+The `json-schema` gem is slow compared to, for example, [JSV](https://github.com/garycourt/JSV). Setting the `--no-validate` switch and running JSON Schema validations separately can further reduce a scraper's running time.
+
+### Profiling
+
+You can profile your code using [perftools.rb](https://github.com/tmm1/perftools.rb). First, install the gem:
+
+    gem install perftools.rb
+
+Then, run your script with the profiler (changing `/tmp/PROFILE_NAME` and `script.rb` as appropriate):
+
+    CPUPROFILE=/tmp/PROFILE_NAME RUBYOPT="-r`gem which perftools | tail -1`" ruby script.rb
+
+You may want to set the `CPUPROFILE_REALTIME=1` flag; however, it seems to change the behavior of the `json-schema` gem, for whatever reason.
+
+[perftools.rb](https://github.com/tmm1/perftools.rb) has several output formats. If your code is straight-forward, you can draw a graph (changing `/tmp/PROFILE_NAME` and `/tmp/PROFILE_NAME.pdf` as appropriate):
+
+    pprof.rb --pdf /tmp/PROFILE_NAME > /tmp/PROFILE_NAME.pdf
+
+## Testing
+
+**DO NOT** run this gem's specs if you are using Redis database number 15 on `localhost`!
+
 ## Bugs? Questions?
 
 This project's main repository is on GitHub: [http://github.com/opennorth/pupa-ruby](http://github.com/opennorth/pupa-ruby), where your contributions, forks, bug reports, feature requests, and feedback are greatly welcomed.
 
 Copyright (c) 2013 Open North Inc., released under the MIT license