Recently in Cache System Category

Web cache is very important to a high traffic site, by holding those pictures, css and js files those are not changed frequently, it can reduce pressure to the backend servers, it's very helpful to improve the site speed.

As you may know, Squid is widely used for web cache service, but we have another choice, that is Varnish. Here is the introduction about Varnish from it's offical site:

"Varnish is a state-of-the-art, high-performance HTTP accelerator. Varnish is targeted primarily at the FreeBSD 6/7 and Linux 2.6 platforms, and takes full advantage of the virtual memory system and advanced I/O features offered by these operating systems."

Some of my friends have already implemented Varnish as their web cache server in the production platform, and most of them replaced Squid with Varnish, they told me that Varnish has better performance than Squid in the high traffic site.

So I wanna install a web cache server for my site, although my site traffic is pretty low :)

Firstly, I'd like show you the basic workflow about how web cache works with web server.


 


Ok, let's start to install Varnish, the last stable version of Varnish is 1.1.2, you may download the source code from here. My current web server is running on FreeBSD, Apache, PHP and MySQL, and Varnish will be installed on RedHat Linux.

Installation
1. Createing group and user 'www' for Varnish
# groupadd www
# useradd www -g www

2. Compiling and installing Varnish
# cd varnish-1.1.2
# ./configure --prefix=/home/varnish
# make && make install

3. Creating cache file and log file folder
# mkdir -p /data02/cachefile
# chown -R www.www /data02/cachefile
# mkdir -p /data02/cachelog
# chown -R www.www /data02/cachelog

4. Varnish configuration file
Example from my site:

backend mywebserver {
       set backend.host = "YOUR_WEB_SERVER_IP";
       set backend.port = "80";
}

acl purge {
       "localhost";
       "127.0.0.1";
       "192.168.1.0"/24;
}

sub vcl_recv {
       if (req.request == "PURGE") {
               if (!client.ip ~ purge) {
                       error 405 "Not allowed.";
               }
               lookup;
       }

       if (req.http.host ~ "^www.isoracle.com") {
               set req.backend = mywebserver;
               if (req.request != "GET" && req.request != "HEAD") {
                       pipe;
               }
                 elseif(req.url ~ "\.(php|cgi|pl)($|\?)") {
                       pass;
                 }
               else {
                       lookup;
               }
       }
       else {
               error 404 "isoracle.com Cache Server";
               lookup;
       }
}

sub vcl_hit {
       if (req.request == "PURGE") {
               set obj.ttl = 0s;
               error 200 "Purged.";
       }
}

sub vcl_miss {
       if (req.request == "PURGE") {
               error 404 "Not in cache.";
       }
}

sub vcl_fetch {
       if (req.request == "GET" && req.url ~ "\.(txt|js)$") {
               set obj.ttl = 3600s;
       }
       else {
               set obj.ttl = 10d;
       }
}

For more detailed description about the Varnish configuration file, please refer to the offical document.

5. Starting varnishd
# /home/varnish/sbin/varnishd -n /data02/cachefile -f /home/varnish/vcl.conf -a 0.0.0.0:80 -s file,/data02/cachefile/cache.dat,500M -g www -u www -w 50,200,10 -T 127.0.0.1:1800 -p client_http11=on

For more options about starting up varnishd, please refer to "varnishd --help".

6. Starting varnishncsa
# /home/varnish/bin/varnishncsa -n /data02/cachefile -w /data02/cachelog/varnish.log &

7. Checking Varnish status
# /home/varnish/bin/varnishstat
This command will show you the current Varnish status, such as cache hints, cache misses etc, they're useful for health check.

Verification
We can use curl to check the HTTP header information:

# curl --head http://www.isoracle.com/
HTTP/1.1 200 OK
Server: isoracle/Stable 1 (Unix) PHP/5.2.0
Last-Modified: Sat, 12 Jul 2008 21:54:39 GMT
ETag: "96347-cc05-4879281f"
Content-Type: text/html
Content-Length: 52229
Date: Sun, 13 Jul 2008 01:36:12 GMT
X-Varnish: 1652882410 1652882077
Age: 8877
Via: 1.1 varnish
Connection: keep-alive

We can see that Varnish works now.

Finally, you need change your DNS record to point the web server IP to cache server IP to let the requests hint the cache server.

Memcached is a great caching application which is widely used in many web 2.0 site, such as LiveJournal, Facebook, Sourceforge etc. I'm also very interest in this software, so first step, let's try to install and use it. For more information about Memcached, please refer to http://danga.com/memcached/

Here, I'd like show you how to install Memcached on RedHat Advanced Server Update 6.

Dependencies

• libevent, http://www.monkey.org/~provos/libevent/
• Linux kernel 2.6, we need a kernel with epoll.

Installation:

• libevent

./configure --prefix=/usr
make && make install

• memcached

There have many configuration options to use when you configure memcached, all the support options are available from:
./configure --help

./configure --prefix=/home/memcached --with-libevent=/usr
make && make install

Starting memcached
/home/memcached/bin/memcached -d -m 100 -l 192.168.1.136 -p 11211 -u nobody

-d: run as daemon
-m: how many memory in size of MB to allocate to memcached
-l: the listen on address which running the memcached
-p: the port allocate for memcached
-u: the user which running memcached daemon

After starting memcached daemon, we can use 'netstat' to verify the port 11211 is opened.
#netstat -an
Active Internet connections (servers and established)
Proto Recv-Q Send-Q Local Address               Foreign Address             State
tcp        0      0 192.168.1.136:11211         0.0.0.0:*                   LISTEN

Some simple Perl scripts to test memcached

We have installed and run the memcached on 192.168.1.136, next step, I'll do some tests on another box 192.168.1.137 to push/pull the data from 192.168.1.136.

Before run the Perl testing scripts, we need install 'Cache::Memcached' module, find it from cpan. Then let's run the following two scripts on the client 192.168.1.137.

1. Pushing the data to memcached
Name: push_mem.pl

#!/usr/bin/perl
use Cache::Memcached;
my $memd = new Cache::Memcached {
      'servers' => [ "192.168.1.136:11211"],
    };

# Set a value
$memd->set("my_key", "123");

$memd->disconnect_all();
exit;

2. Pulling the data from memcached
Name: print_mem.pl

#!/usr/bin/perl
use Cache::Memcached;
my $memd = new Cache::Memcached {
      'servers' => [ "192.168.1.136:11211"],
    };
my $val = $memd->get( "my_key" );
if ( $val )
{
     print "Value is '$val'\n";
}

$memd->disconnect_all();
exit;

Result: Value is '123'

Ok, we can see that the value '123' was stored in memcached when I run the 'push_mem.pl' for the first time, and then we get the same value from memcached when I run 'print_mem.pl'.

Here, I only show you the the simplest examples on how to use memcached, but in a production environment, that will be more complex.

Tips:
We can telnet memcached with port 11211 to check the memcached status, it will be helpful for troubleshooting.

[root@web1 scripts]# telnet 192.168.1.136 11211
Trying 192.168.1.136...
Connected to web1.isoracle.com (192.168.1.136).
Escape character is '^]'.
input command: stats
STAT pid 23810
STAT uptime 622
STAT time 1209240944
STAT version 1.2.5
STAT pointer_size 32
STAT rusage_user 0.000999
STAT rusage_system 0.061990
STAT curr_items 1
STAT total_items 1
STAT bytes 58
STAT curr_connections 2
STAT total_connections 5
STAT connection_structures 3
STAT cmd_get 1
STAT cmd_set 1
STAT get_hits 1
STAT get_misses 0
STAT evictions 0
STAT bytes_read 43
STAT bytes_written 36
STAT limit_maxbytes 104857600
STAT threads 1
END

And we can calculate the hint rate, the hint rate= get_hits/ cmd_get