Why Bootdev — Dynamic CDN

In the old days, we put images, css / js, woff etc any assets to CDN, so that clients can download somewhere that geographically optimised.

Around end of 2012 – early 2013, new idea comes out like we should CDN everything, so that we can reduce the complex architecture with memcache, page cache cluster (Varnish) or even microccache etc. Just 1 layer cache and having everything to CDN. Like architecture below.

dnamic cdn

Your website domain name will directly point to CDN with CNAME. And then the CDN will point to your load balancer address or your web server. So it is like a proxy. When you do create, it will 100% bypass CDN and goes to web server, when u do UPDATE, the CDN will update from web server then invalidate itself, when you do DELETE, the CDN will bypass request to web server and invalidate its cache. When you read, it read from the CDN, not from web server. So the CRUD action can be completed.

You will need your own cache invalidation strategy, like send update to cloudfront / using versioning object or url.

Here is a sample conf of how we bypass some URLs to go to web server, and making Drupal works.

E1EB6A9C-17EC-446D-AD59-80B471A4F962 62367506-DDC3-4E5C-8F05-24E2D20DBBBB

With AWS cloudfront, you can bypass header ORIGIN, so that you can preform CORs actions. Also you can use similar header bypass feature to detect mobile/PC. With such architecture well setup, theoretically, you can have unlimited PV, as your server wont be really hitted. Your bound will be write DB bound only, which is not a concern in most case.

If you don’t want to understand all these, but want to lower your cost and have higher traffic and faster response, contact bootdev at founders@bootdev.com ! We can deploy Dynamic CDN to you in minutes no matter you are using AWS or not. We can point our CloudFront account to your server, it can be Azure, Linode, or any bare-meter. It just need to be Drupal, and you can enjoy the best performance ever.

ref: https://media.amazonwebservices.com/blog/cloudfront_dynamic_web_sites_full_1.jpg

Advertisements

why BootDev — AWS management (EC2 decrease performance by time)

In our experience, if your EC2 got resource left for a period of time, lets say 1-2 months, even AWS say it is not, but we always experience performance drop. For example: server response time drop from 400ms to 600ms . Or CPU raised from 40% to 70% in average.

In many case, many users will not 100% * 24 * 7  consume their server resources. So, in my experience, AWS want to reduce cost and reallocate some resource privately to other clients. If your server is actually low utilization, it is OK. But, in my situation, we need high response time and we reserve the CPU for that. We will feel performance drop.

How we deal with that ? That is what auto-scaling group should do. If in your auto-scaling group, your server will keep renew itself. Like turning off old servers and start new servers, the EC2 resource allocation will maintain at best performance.

Such issue especially obvious in Drupal. Drupal PHP request involve in many hooks / functions / third party calls etc. Server CPU requirement is higher than other web PHP applications. So, renewing server can help to keep better performance.

How to deploy auto-scaling group ? Call BootDev 🙂

Hope above information can help anyone that feel AWS sometimes drop performance by time 🙂

 

 

 

 

Why bootdev — Nginx conf extended

Recently, one of our Drupal stack really high CPU, It do monthly increased 30% RPM (request per minute) from new relic. BTW, i think its CPU wont increase like 40%. During my investigation,

  1. some robots scan can somehow eat your CPU. Like many robots will look for wp-login as it guess you are wordpress site and more.
  2. Your CDN may pull your data but goes to a wrong URL and have a ERROR in your php-fpm log
  3. robot from same IP address(es) keep scan your site

We here added some conf which help about 6-7% CPU decrease for a 300-400 RPM Drupal site.

#Set a limit connection zone
limit_conn_zone $binary_remote_addr zone=gulag:5m;

#Check robot
map $http_user_agent $is_bot {
default  ”;
~*crawl|goog|yahoo|yandex|spider|bot|tracker|click|parser is_bot;
}
###
### Send all known bots to $args free URLs.
###
location @nobots {
if ($is_bot) {
rewrite ^ $scheme://$host$uri? permanent;
}
rewrite ^/(.*)$  /index.php?q=$1 last;
}

#In CDN Config and Advagg conf of Drupal nginx,
#add below to restrict same IP only allow 32 connections
limit_conn gulag 32;
#Block connection if it is robots (Put anywhere u like)
try_files  $uri @nobots;

#Block wordpress scan attack
location ~ ^/(wp-admin|wp-login\.php) {
deny all;
}

The block wordpress scan thing can add anything that u find strange from php-fpm/error.log by just changing the path pattern

I have load balancer setup with 2 web server, one with and one without config above. The different on Amazon m3.large is about 6% in a 300-400RPM website, in which both server throughput the same RPM.

Of coz, the above settings is updated into BootDev Nginx conf.

Enjoy 🙂

Why BootDev — Nginx Config

We had spent many effort on Nginx configuration. Different with Apache + mod_php5, Nginx + php_fpm need much detail configuration and nginx is Drupal module dependent. It means some Drupal module require support from Nginx configuration. Like CDN module / Advagg module.

There is a github about Drupal + Nginx, but that will be too much and you will require to filter the necessary part in your project.

Here i share the main nginx configure

server {
  server_name *.compute.amazonaws.com;
  root   /opt/source/app;
  access_log  /var/log/nginx/access.log;
  error_log  /var/log/nginx/error.log;

  #include /etc/nginx/apps/drupal/drupal.conf;
  #Cache everything by default
  set $no_cache 0;
  #Don't cache POST requests
  if ($request_method = POST)
  {
    set $no_cache 1;
  }

  #Don't cache if the URL contains a query string
  if ($query_string != "")
  {    
    set $no_cache 1;
  }

  #Don't cache the following URLs
  if ($request_uri ~* "/(administrator/|login.php)")
  {
    set $no_cache 1;
  }

  #Don't cache if there is a cookie called PHPSESSID
  if ($http_cookie = "PHPSESSID")
  {
    set $no_cache 1;
  }

  # Enable compression, this will help if you have for instance advagg module
  # by serving Gzip versions of the files.
  gzip_static on;

  location = /favicon.ico {
    log_not_found off;
    access_log off;
  }

  location = /robots.txt {
    allow all;
    log_not_found off;
    access_log off;
  }

  # This matters if you use drush
  location = /backup {
    deny all;
  }

  # Very rarely should these ever be accessed outside of your lan
  location ~* \.(txt|log)$ {
    deny all;
  }

  location ~ \..*/.*\.php$ {
    return 403;
  }

  location / {
    # This is cool because no php is touched for static content
    try_files $uri @rewrite;
  }

  location @rewrite {
    # Some modules enforce no slash (/) at the end of the URL
    # Else this rewrite block wouldn't be needed (GlobalRedirect)
    rewrite ^/(.*)$ /index.php?q=$1;
  }

  location ~ \.php$ {
    fastcgi_split_path_info ^(.+\.php)(/.+)$;
    #NOTE: You should have "cgi.fix_pathinfo = 0;" in php.ini
    include fastcgi_params;
    fastcgi_param SCRIPT_FILENAME $document_root$fastcgi_script_name;
    fastcgi_intercept_errors on;
    fastcgi_pass unix:/var/run/php-fpm-www.sock;
    fastcgi_read_timeout 40;
    fastcgi_cache MYAPP;
    fastcgi_cache_valid 200 301 30s;
    fastcgi_cache_bypass $no_cache;
    fastcgi_no_cache $no_cache;

    # Set cache key to include identifying components
    fastcgi_cache_valid 302     1m;
    fastcgi_cache_valid 404     1s;
    fastcgi_cache_min_uses 1;
    fastcgi_cache_use_stale error timeout invalid_header updating http_500;
    fastcgi_ignore_headers Cache-Control Expires;
    fastcgi_pass_header Set-Cookie;
    fastcgi_pass_header Cookie;

    ## Add a cache miss/hit status header.
    add_header X-Micro-Cache $upstream_cache_status;

    ## To avoid any interaction with the cache control headers we expire
    ## everything on this location immediately.
    expires epoch;

    ## Cache locking mechanism for protecting the backend of too many
    ## simultaneous requests.
    fastcgi_cache_lock on;
  }

  # Catch image styles for D7.
  location ~ ^/sites/.*/files/ {
    try_files $uri @rewrite;
  }

  # Catch image styles for AmazonS3 D7.
  location ~ ^/system/files/styles/ {
    try_files $uri @rewrite;
  }

  location ~* \.(js|css|png|jpg|jpeg|gif|ico|svg)$ {
    expires max;
    log_not_found off;
  }

  location ~* \.(eot|ttf|woff|svg) {
    add_header Access-Control-Allow-Origin *;
    try_files $uri @rewrite;
  }

  ##   
  # Advanced Aggregation module CSS
  ##   
  # http://drupal.org/project/advagg.
  ##    
  location ^~ /sites/default/files/advagg_css/ {
    expires max;
    add_header ETag '';
    add_header Last-Modified 'Wed, 20 Jan 1988 04:20:42 GMT';
    add_header Accept-Ranges '';
    add_header Access-Control-Allow-Origin *;
    
    location ~* /sites/default/files/advagg_css/css__[[:alnum:]-_]+\.css$ {
      access_log off;
      try_files $uri @drupal;
    }
  }

  ###
  ### CDN Far Future expiration support.
  ###
  location ^~ /cdn/farfuture/ {
    tcp_nodelay   off;
    access_log    off;
    log_not_found off;
    etag          off;
    gzip_http_version 1.0;
    if_modified_since exact;
    location ~* ^/cdn/farfuture/.+\.(?:css|js|jpe?g|gif|png|ico|bmp|svg|swf|pdf|docx?|xlsx?|pptx?|tiff?|txt|rtf|class|otf|ttf|woff|eot|less)$ {
      expires max;
      add_header X-Header "CDN Far Future Generator 1.0";
      add_header Cache-Control "no-transform, public";
      add_header Last-Modified "Wed, 20 Jan 1988 04:20:42 GMT";
      rewrite ^/cdn/farfuture/[^/]+/[^/]+/(.+)$ /$1 break;
      try_files $uri @nobots;
    }
    location ~* ^/cdn/farfuture/ {
      expires epoch;
      add_header X-Header "CDN Far Future Generator 1.1";
      add_header Cache-Control "private, must-revalidate, proxy-revalidate";
      rewrite ^/cdn/farfuture/[^/]+/[^/]+/(.+)$ /$1 break;
      try_files $uri @nobots;
    }
    try_files $uri @nobots;
  }

}

The idea of this config file is to support CDN far future, CDN, advagg, Drupal image styling, AmazonS3 and microcache modules. You need to catch different url pattern for different purpose.

For microcache, we put the cache into memory and expire every 30s, so that in each 30s, only the 1st visitor hit your site will generate by PHP. 2nd – N users will hit microcahe. In this approach, we can support high traffic website and then same time avoid handling cache invalidation problem.

Here I also share the cache config which we put nginx cache into memory which release better performance.

fastcgi_cache_path /dev/shm/microcache levels=1:2 keys_zone=MYAPP:5M max_size=256M inactive=2h;
fastcgi_cache_key "$scheme$request_method$host$request_uri";
add_header X-Cache $upstream_cache_status;
map $http_cookie $cache_uid {
  default nil; # hommage to Lisp :)
  ~SESS[[:alnum:]]+=(?<session_id>[[:alnum:]]+) $session_id;
}
map $request_method $no_cache {
  default 1;
  HEAD 0;
  GET 0;
}

You can read the comment inside the config file for more detail explanation.

This config requires to work with another PHP-FPM config, so that the memory is optimized. Then, you can estimate how many request per second that your server can serve. And i will talk about it next time.

Linode vs AWS pricing

Recently, i moved my linode to new plans and compare it with latest EC2 pricing. I have myself 2 “1024” and 1 “2048” server in linode which just cost USD45 per month. Personally which is a very good deal. I served my clients on AWS as AWS can do much more things and actually cheaper if you buy bigger servers.

It is actually not easy to find a better price between them if you have no data on your application. In my application, bottleneck is on Memory instead of CPU. I believe most application are same with me, that is why Linode pricing plan base on Memory as well.

Monitoring

If your server shows CPU always 10-20% and you are running websites / PHP things. This price comparison may be useful to you.

Memory

Screen Shot 2014-12-17 at 2.11.21 PM Screen Shot 2014-12-17 at 2.10.38 PM

Comparing Linode 8192 plan and EC2 m3.large, the monthly price is AWS more expensive. But, dont forget Linode require USD20 extra for the backup plan, so it will be same price as AWS. Someone may say AWS lot fewer computation power, i agree with it. My application is memory driven, so i don’t care.

In my case, AWS is cheaper when the machine is more powerful. Like if we just need 1024 / 2048 memory, USD 10/20, you can buy nothing in AWS, but you can have really good servers at linode.

Extra features

While price don’t change, AWS also offer VPC, cross region AMI copying, IAM user control, load balancer pay by traffic, nearly free S3 storage which you will be expensive to have all these in Linode.

Availability

One of the problem of Linode is their servers will sold out. So, for enterprise usage, like ElasticMapReduce / Auto-scaling, you cannot do it on Linode. Remember, High availability is cost. AWS reserved many instance in each data center, so that you can pay for it when you need it. Linode cannot offer like if I need 10 more servers anytime I want.

Summary

In Summary, i would suggest people to use Linode if you are just considering to have < 4096 memory per server. You dont need architecture like separate servers into NAT zones.

If you are serving big server architecture, you are better to go for AWS due to rich features and availability. And for the big servers, AWS actually 50%+ cheaper when you go for the pre-paid reserved instances.