Apache Traffic Server

Increase uptime and performance with Apache Traffic Server Tom Melendez, Yahoo!

About Tom Did Infrastructure development at Y! News for 3 years; have since moved on to do non-infra Hadoop stuff General philosophy: My job isn’t done until I’m not doing it anymore What I’ve learned: If Devs want to be happy, make Ops happy. Making Ops happy will ultimately make the business happy. Primary motivation in life: I only do the fun stuff.

Agenda What Who Where and When: The Internet. Now. Why/How

What Very efficient caching proxy. Multi-threaded, event driven I’ve seen it handle 30k rps per box (quad-core) without issue I’ve seen lab data quoting 300k+ rps per box Delivers 400 terabytes of data per day at Y! Probably a configuration that can help your app.

Who (plus some history) Open-Sourced by Yahoo! in November 2009 Yahoo! received it with Inktomi acquistion Yahoo! currently has a team of engs on it We use it all over the place Apache TLP Core committers from Yahoo!, Google, Akamai and elsewhere

Where and When The internet. Now.

Why? You want to build a CDN You want a reverse proxy You want a forward proxy You want good response for your worldwide customers If you’re like me, you don’t want to do a lot of work. 

How? - Requirements Well, you need to be able to compile the source. (So, Linux, OSX or Solaris) There are also instructions for installing on EC2 as well as AMIs available

How - installation ./configure && make && make install Depending on your time to market, please consider working with the developer release, otherwise use stable Start it up sudo /usr/local/trafficserver start But you’ll probably want to configure it first. 

How - executables Processes: traffic_server – handles the transactions traffic_manager – manages the traffic server traffic_cop – health check for the above Tools: traffic_line – gets stats and (re)read config

How – the config files bypass.config - static bypass rules that Traffic Server uses in transparent proxy caching mode. cache.config - how Traffic Server caches web objects. congestion.config - enables you to configure Traffic Server to stop forwarding HTTP requests to origin servers when they become congested, filter.config - enables you to deny or allow particular requests and strip header information from client requests. hosting.config - assign cache partitions to specific origin servers icp.config - defines ICP peers (parent and sibling caches). ip_allow.config - controls client access to the Traffic Server proxy cache.

How – the config files logs.config - formats traditional custom transaction log files. log_hosts.config - log transactions for different origin servers in sep. files logs_xml.config - defines the custom log file formats, filters, and processing options. parent.config - identifies the parent proxies partition.config - enables you to manage your cache space by creating partitions records.config - list of configurable variables used by the Traffic Server software.

How – the config files remap.config - contains mapping rules that Traffic Server uses to perform various actions. splitdns.config - file enables you to specify the DNS server that Traffic Server should use for resolving hosts under specific conditions. ssl_multicert.config - configure Traffic Server to use multiple SSL server certificates with the SSL termination option. storage.config - lists all the files, directories, and/or hard disk partitions that make up the Traffic Server cache. update.config - how Traffic Server performs a scheduled update of specific local cache content.

How - configuration There are a lot of configuration files I know what you must be thinking….

But remember… I’m really lazy.

How - configuration Never fear. You primarily deal with this one (routing rules): remap.config You’ll tune this one when you’re first setting it up (system vars): records.config If you’re using caching, you’ll set up this one: storage.config

How – config – Reverse Proxy ATS acts as the origin server IMO, this is the major use case ATS can front one more more websites

How – config – Reverse Proxy records.config: CONFIG proxy.config.proxy_name STRING <hostname> CONFIG proxy.config.http.server_port INT 80 CONFIG proxy.config.reverse_proxy.enabled INT 1 Depending on what you’re in front of: CONFIG proxy.config.http.keep_alive_no_activity_timeout_in INT 60 CONFIG proxy.config.http.keep_alive_no_activity_timeout_out INT 1 CONFIG proxy.config.http.transaction_no_activity_timeout_in INT 15 CONFIG proxy.config.http.transaction_no_activity_timeout_out INT 30 CONFIG proxy.config.cache.ram_cache.size LLONG 1332735284

How – config – Reverse Proxy storage.config /dev/raw_sdb 1078000000 NOTE: To use RAW disk you must use the '/usr/bin/raw' program to bind a raw device to a an existing block device.

How – config – Reverse Proxy remap.config map http://www.example.com http://realsite.example.com Now, reload traffic server: sudo /usr/local/bin/traffic_line -x

How - quick notes about cache You can purge objects from the cache with PURGE curl -X PURGE -v http://example.com/remove_me.jpg There is a UI to inspect and modify the cache. You can enable it with: CONFIG proxy.config.http_ui_enabled INT 1 (I came to learn very recently that the UI won’t be supported, so if you really want it, make a case for it)

How – config – General Proxy Forward Proxy Your FEs make requests to APIs (yours and partners/external) You don’t need to keep a sep proxy technology around (i.e. Squid), ATS does this too.

How – config – General Proxy You can do all the cool cache stuff configuration for computation for object freshness Push content into the cache Pin contents in the cache for specific periods Ignore no-cache headers Translation of Squid directives to ATS: https://cwiki.apache.org/confluence/display/TS/SquidConfigTranslation

But, but, but…. “ Well, I don’t expect lots of traffic” “ I already have VIPs/Load Balancers” “ Do I really need a proxy server?” “ Is this yet another thing I need Ops for?”

How – Usage Even if you can handle your traffic: change origin servers without changing DNS redirect different URL paths, or rewrite URLs altogether send traffic cross-colo during upgrades, for testing, etc. use minimal server installations across the world, just put ATS there and have DNS resolution point there.

Extending How - Plugins You can extend the functionality of ATS with plugins There are plugins available (forthcoming) by Yahoo! You can build your own plugins and there is lots of documentation available. You are STRONGLY encouraged to contribute back!

Example functionality of plugins Modify the response Do a db lookup based on a header append or remove query string params or headers before they are sent to the origin server Respect custom headers More advanced URL rewriting (more than regex)

Ok, nothing is that easy Obviously, create the remap.config and rules You may want to create your own plugins (probably not) Code changes If you plan on using the cache, your cache control headers should be configurable (you should be doing this anyway) Process changes Ops should own YTS, dev might need to set it up QA needs to have YTS in their env; devs should have one too Ops changes If you don’t have monitoring, this would be a good time to add it.

Competitors There are plenty of Open Source competitors Not easy to make an apples-to-apples comparison I don’t have any benchmarks Look for the features you want and an active community.

Competitors: A stolen slide, probably not up-to-date! ATS HAproxy nginx Squid Varnish mod_proxy Worker Threads Y N N N Y Y Multi-Process N Y Y N Y Y Event-driven Y Y Y Y N N? Plugin APIs Y N Y part Y Y Forward Proxy Y N N Y N Y Reverse Proxy Y Y Y Y Y Y Transp. Proxy Y Y N Y N N Load Balancer part Y Y Y Y Y Cache Y N Y Y Y Y ESI soon N N Y Y N ICP Y N N Y N N Keep-Alive Y N Y Y Y Y SSL Y N Y Y N Y

What is ESI? “ Edge Side Includes” Think SSI on the “Edge” Example in your markup (taken from Wikipedia) <esi:include src="http://example.com/1.html" alt="http://bak.example.com/2.html" onerror="continue"/> (Something to think about – ESI will change your development processes, testing, etc.)

ESI Another Example <esi:try> <esi:attempt> <esi:comment text="Include an ad"/> <esi:include src="http://www.example.com/ad1.html"/> </esi:attempt> <esi:except> <esi:comment text="Just write some HTML instead"/> <a href=www.akamai.com>www.example.com</a> </esi:except> </esi:try> Read the spec at: http://www.w3.org/TR/esi-lang

Questions? http://trafficserver.apache.org/docs/v2/admin/ http://ostatic.com/blog/guest-post-yahoos-cloud-team-open-sources-traffic-server https://cwiki.apache.org/confluence/display/TS/RoadMap http://www.oscon.com/oscon2010/public/schedule/detail/13878

Questions for the Audience (you!) I have some of my own: Developing in something other than PHP? Anyone moving away from PHP? If so, why? Big challenges you’re facing right now? Frameworks/CMSes – not a lot of talk about them this year What technology changes are you investing in? Are you doing unit/functional/perf testing regularly?

Apache Traffic Server

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Apache Traffic Server (20)

Recently uploaded (20)

Apache Traffic Server

Editor's Notes