WordPress is one of today’s most popular CMS/blogging platforms. Millions of websites rely on its simplicity, its wealth of capabilities and its plug-ins to manage their online presence and to ensure their website is available 24x7x365.

Although most sites don’t require a huge amount of capacity (and thereby require only a single server), high traffic sites (like several of our customers) often split traffic across multiple high-end servers, which requires their WordPress setup to be optimized and perfect: every little detail matters.

So what can you do to make sure your WordPress site runs efficiently and utilizes fewer resources?  One option is to hire dozen of developers (and spend hundreds of thousands of dollars) to optimize your code, which will result in an environment that doesn’t really look like WordPress anymore. You probably won’t be able to take advantage of the thousands of free plugins and, most importantly, you probably won’t be able to get WordPress updates (because your code is too customized). Perhaps not the best option.

A second (better) option would be to improve the underlying technology that powers WordPress: the web server, the PHP interpreter and the database server. This will allow you to continue reaping the benefits of WordPress updates and plugins, with the added benefit of having an optimized, high performance environment.

With the second option in mind, Voxel has spent several weeks developing an optimized WordPress environment that we’re now ready to share with you. What follows is the software stack that we have included (and we think it’s da BOMB).

NGINX

For web server software, we’re using NGINX. NGINX (pronounced “engine-x”) is open source web server software that is designed to provide high performance, low memory usage and high concurrency. The software is absolutely stunning – it uses a fraction of the resources other software packages (like Apache) use to deliver requests, and delivers them FASTER. On top of all that, the configuration is so simple that anyone can do it.

There are tons of references on the web about NGINX’s legendary performance, and their official English site has a great page with quotes from people who are using it for heavily loaded environments: Why Use NGINX?

Some big names out there that are leveraging the power of NGINX are: WordPress, Hulu, Github, Ohloh, SourceForge,WhitePages,TorrentReactor – just to name a few.

PHP-FPM (with APC)

PHP-FPM (FastCGI Process Manager) is what we’re using as a PHP interpreter.  PHP FastCGI implementations provide a number of extra features that are especially beneficial for heavily loaded web servers. Initially it was released as a set of patches for specific PHP stock versions but since PHP version 5.3.3 it has become part of the official PHP distribution.
WordPress PHP code won’t notice a difference (and will run more smoothly).

We have also included the well-known APC in the mix, which is a robust open source framework for caching and optimizing PHP intermediate code.

PERCONA SERVER

The Percona Server is an enhanced drop-in replacement for MySQL server.  It includes an XtraDB storage engine and provides extra features, better performance and improved instrumentation for analysis of performance and usage.

The Percona Server people are also responsible for the popular MySQL Performance Blog and some other papers and tools related to MySQL.

This particular environment is not available yet but will be available soon (after we’ve completed thorough testing).  You’ll be able to provision it through our hAPI for VoxCLOUD and VoxSERVERS. The provisioning will automatically adjust your configurations to fit the server’s specs.

If you are using (or planning to use) Chef this will suit you perfectly.  Why? Because we have Chef-ized the environment allowing you to provision it using Knife.  You can read more about it James Brinkerhoff’s blog post: Bring Your Knife to Voxel’s hAPI Kitchen.

I fly a lot for work and generally have had very positive on-time and trouble-free experiences. It just so happens that recently, I fell victim to an equipment failure and lack of planning that ultimately brought me back to earth about how these issues seem to plague the transit industry. Most travelers focus on the impact to their lives and plans without thinking of the cascading effects including missed connections and schedules of all other flights dependent that day on this particular plane.

I contacted a co-worker and mentioned the issue and he joked that I should have had a redundant ticket as a “plan B.” My response was “Why put the burden on the business customer for redundancy when the airline should have a responsibility for keeping spare parts and planes?” In the real world failure is not an option, it’s a reality. Businesses that don’t have redundancy will eventually find their customers going to a competitor whose services are available all the time.

Redundancy in most real world cases may seem like a bad idea because it requires you to invest resources into non-essential components. This is not to say that all aspects of business need to be fully redundant, but each aspect should be evaluated for risks where failures can occur. Minimizing costs and maximizing coverage to prevent a critical service failure is the overall goal. In the managed hosting space, identifying the components most likely to fail and bring a customer’s solution to a halt is crucial.

If downtime is not an option, a redundant solution, or N+1, may be required. This means that one component at each level of a solution is a redundant copy and the solution can withstand a failure of one device at each level. This configuration allows a customer to have the assurance of being able to withstand a single failure or multiple failures at different levels without it impacting the integrity of the business.

As an IT Infrastructure provider, we are able to maintain an inventory of these components on hand to ensure maximum uptime…too bad my airline can’t do the same thing.

The discussion began something like this:

Me: I think more people would read our blog if it was more fun.
Marketing: We are blogging about IT infrastructure. How funny is that???
Me: I didn’t say funny. I think this stuff can be fun. Maybe even a bit funny too.
Marketing: Even if you could make it funny, you’d never get it past Legal.

So I worked up a joke, based on the funniest joke in the world as reported by CNN Tech. And I got it past Legal. Here’s how it goes:

Two Systems Administrators are out in the data center installing servers and excitedly discussing Internap’s new Colocation ebook when one of them collapses. He doesn’t seem to be breathing and his eyes are glazed. The other guy screams for a Facility Engineer who comes running over.

The Sys Admin guy gasps: “My friend is dead! What can I do?” The Facility Engineer says: “Calm down, I can help. First, let’s make sure he’s dead.” The Sys Admin grabs a server, smashes his friend on the head and turns to the Facility Engineer and says: “OK, now what?”

Did you notice that I managed to incorporate a reference to our new colocation ebook? It does a really nice job of helping you understand if colocation is right for your business. If you’ve ever wondered how colocation can benefit your business, we lay it out pretty clearly in terms of risk mitigation and cost savings. Rather eye opening, mind expanding stuff for those who are yet colo people. It’s kinda like being hit in the head with a server. Without the death.

IT Infrastructure. Check. Fun. Check. Funny. Check. Legal approval. Check. Bonus points for containing relevant content. Check.

What did I experience at the OpenStack Essex Design Summit in Boston?

I was at the OpenStack Essex Design Summit in Boston last week from Oct. 13–15 (we’re making a commitment to OpenStack here at Voxel), and here are some of my adventures and lessons learned.

Why does OpenStack feel like the next Apache?

OpenStack, the open source cloud services software suite, has wind beneath its wings. Even though I only focused on one specific project last week, if the other projects have half as much momentum, smarts, and good-natured collaboration, then all of them are sure to succeed. OpenStack could possibly have the same effect that the Apache web server had on Internet infrastructure.
I remember the early days of the ‘Net, when I’d check Netcraft.com to find out which web server was leading the “most sites hosted” competition. I was working for Netscape at the time, and we were always hungry to steal rank from Microsoft. Apache was too formidable an opponent, so we resigned ourselves to taking market share from the other “enterprise” web servers. Although we wanted to penetrate the service provider market, our features were too enterprise-y and not service provider-y enough. Plus, who can compete with free and open source?

What kind of competition is emerging in the cloud ecosystem?

I see the same sort of (mostly) friendly competition unfolding on the cloud and DevOps mailing lists today. What types of companies will choose which type of cloud service software? Smart shops look carefully at feature set, operational complexity, and scalability goals. They also study the other organizations committing to the project—the quality of the developers and the tone on the mailing lists and IRC. So far, OpenStack has not disappointed.

What makes OpenStack Swift stand out?

I have, admittedly, a somewhat jaded view. I’m focusing on OpenStack’s most successful development project—the Object Storage System called Swift. But throughout all the OpenStack projects I’ve seen, even where quality has been uneven, it’s not for lack of effort. Swift, and all these projects, are moving targets; the drive for greater scalability sometimes causes dramatic design shifts. Still, the organization, layout, and documentation are usable and nimble. At the Summit, the Swift developers were informative, open, and very receptive to feedback from other developers and implementers.

What lessons can OpenStack learn from the Apache Foundation?

I’ve never been to an ApacheCon, but it’s the only group I can think of right now that has the same diversity of products and unity of cause. The Apache project and foundation are highly successful, and I hope the OpenStack Foundation studies Apache’s model closely—learning from its successes and failures alike, not through trial and error but through wisdom and observation.

How was the overall quality of the Summit discussions?

In summary, some talks at the Summit were great, others were unstructured, and some leaders struggled to keep the conversation focused. But that happens at nearly every conference and on most mailing lists. The freedom to digress is often the price of inclusion. If you try to exclude people from the conversation, you risk dull and predictable results. The trick is good leadership—allowing all voices to contribute while still driving toward a goal. And overall, I think OpenStack has strong leadership.

What’s next for me and OpenStack Swift?

I’m now very happily preparing to go to market with OpenStack Swift, encouraged by the steady and humble leadership I witnessed in the OpenStack community and core team. I’m confident that my serious approach to the product will be met with equally serious support when challenges arise—as they inevitably do.

I look forward to your files!

Do original ideas still exist?

If you are like me, you are constantly coming up with what you think are original and innovative ideas. Problem is that they all seem to be taken already! Book concept? Taken. Dot com idea? Taken. Social media idea? Taken! Or wait…is it?

What is Pinterest?

Enter Pinterest. Pinterest is a Virtual Pinboard that lets you organize and share all the beautiful things you find on the web. But most interestingly – you can browse pinboards created by other people, repin their ideas and follow them. Pinterest has quietly been making a huge splash online. Ten thousand avid iPhone app fans? Check. Ten million in funding and a $40 million pre-money valuation? Check. Coaxed me into daily perusing and use of the site? Check!

How does infrastructure support brilliant ideas?

In a perfect world, brilliant ideas (or even just good ones) aren’t limited by technology. When these ideas are being generated on the back of a napkin, my assumption is that IT Infrastructure isn’t top of mind. At the end of the day, after all the planning is done, businesses need a strategy with a complete spectrum of intelligent IT Infrastructure solutions that allow them to grow from a winning idea as a start-up to an enterprise processing millions of transactions.

What should businesses look for in a partner?

Because your business is dynamic and your requirements will evolve, select a partner that can adapt with you – a partner who can support all of your infrastructure needs from colocation to the cloud. Read, De-Risk Your IT Strategy with HorizonIQ’s High-Performance Guarantee to learn how businesses today face extraordinary risk and change while balancing the complexity of enterprise IT and digital innovation.

Is it still a brave new world for ideas?

The good news? It’s still a brave new world. A world where all the good ideas aren’t taken, all the VC funds aren’t dried up and people still want to share their thoughts and ideas with complete strangers. Dream on.

Why is Voxel excited about Chef?

At Voxel, we’re big believers in automating infrastructure, and one of the tools we’re most excited about right now is Opscode’s Chef. Chef provides us with Knife, a convenient tool that allows you to easily manage one or more Chef instances.

How was provisioning handled before?

So, historically, you would use Voxel’s API to provision a VoxCLOUD or VoxSERVER device, then use Knife to bootstrap and spin that provisioned node up as a Chef client. Today we would like to announce that a Voxel specific Knife plugin has been made available via RubyGems/GitHub: ‘knife-voxel’.

What does the knife-voxel plugin do?

Currently on version 0.0.7, the Voxel/Chef plugin allows you to provision (and bootstrap) Voxel’s virtual and dedicated instances (known as VoxCLOUD servers and VoxSERVERS in marketing-speak) as well as list and delete previously provisioned instances.

But there will be a lot more functionality to look forward to in the coming weeks, functionality that will continue to enhance the Voxel/Chef experience.

What commands are available in knife-voxel?

At this time, the knife-voxel plugin has the following commands:

knife voxel voxcloud delete DEVICE_ID (options)

knife voxel voxservers inventory list (options)

knife voxel images list (options)

knife voxel voxcloud create (options)

knife voxel devices list (options)

knife voxel voxservers delete DEVICE_ID (options)

knife voxel voxservers create (options)

knife voxel voxservers reimage DEVICE_ID (options)

These commands mirror Voxel’s hAPI methods as defined @ https://api.voxel.net/docs/.

How is authentication handled?

Authentication can be handled in a few different ways. For each command you can pass –voxel-api-key and –voxel-api-secret with your corresponding hAPI key/secret pair. Or if you prefer, you can utilize your username and password via the –voxel-api-username and –voxel-api-password parameters. Both of these parameter sets can be specified in a knife.rb file, saving you the trouble of entering them each time at the command line. We suggest using a stored authkey/secret pair rather than a username and password when placing these options in knife.rb. The syntax is:

knife[:voxel_api_key] = ‘01234567890’
knife[:voxel_api_secret] = ‘09876543210’

How do you list available images?

To list all available images, and their corresponding IDs:

knife voxel images list

How do you list dedicated VoxSERVER inventory?

To list the available dedicated VoxSERVER inventory, by facility, run the command:

knife voxel voxservers inventory list

How do you provision a VoxCLOUD instance?

To provision a 4 core VoxCLOUD instances with 50GB of disk and Ubuntu10.04 LTS at our LGA8 facility you would run the following:

knife voxel voxcloud create –facility LGA8 –hostname
my.new.chef.instance –processing-cores 4 –disk-size 50 –image-id 55

How do you provision a VoxSERVER?

To provision an available VoxSERVER configuration from inventory with our managed CentOS variant, VSE:

knife voxel voxservers create –facility LGA8 –configuration-id 2
> –hostname my.physical.chef.instance –image-id 16

Where can you find more information?

And if you’re looking for further information for each command issue:

–help

What’s next for knife-voxel?

We’re really excited to be able to offer these infrastructure automation tools to our clients – look for more to come!

What are my thoughts from the Fall 2011 OpenStack Design Summit and Conference?

Thoughts from the Fall 2011 OpenStack Design Summit and Conference

As the conference is coming to a close in Boston today, I wanted to provide two observations.

Why is Rackspace’s announcement about the OpenStack foundation so significant?

First, the announcement by Rackspace this week that they would be turning over the intellectual property and trademarks for OpenStack to a not-for-profit foundation in 2012 marks a potentially watershed moment for the future of OpenStack.

One of my main motivations for hopping a flight to Boston earlier this week was to get a sense for the future of governance and strategy for the OpenStack™ projects, and I was not disappointed. Despite getting leaked to the media the day before, Rackspace’s formal announcement yesterday was clearly the most impactful development.

It will be up to the members of the OpenStack community to determine the success of this burgeoning open source cloud platform. For sure, this membership is a credible cast and rapidly approaching an ecosystem-level of maturity. Members span a broad swath of cloud service providers like Internap, technology providers, government agencies, universities and research institutions. Some of the biggest names in technology are supporting this effort including Dell, HP and Cisco.

What challenges and opportunities lie ahead for the OpenStack foundation?

As it is formed and control is transferred, creation of the structure and governing rules of the foundation will be a critical activity. The foundation will need to be flexible enough to attract innovative contributions by organizations of all kinds, while protecting the OpenStack brand, maintaining a quality platform and allowing participating organizations to benefit from the shared development.

Done right, OpenStack will make the transition from a growing open source cloud project to a leading platform for the world’s technology needs, which challenges the market power of some of the current proprietary offerings. Done wrong, well … let’s focus on getting it right. To do our part, I look forward to Internap ramping up our participation in the OpenStack community in the coming months.

What opportunities exist for engineers within the OpenStack community?

The show also confirmed that there is a terrific opportunity for innovative engineers in the OpenStack community that is not being fully met.

This week, I met people from around the world who clearly are passionate about the success of OpenStack. This talented group understands what it takes to develop a platform that can scale to massive proportions, deliver on the promise of high availability and provide functionality to support a plethora of cloud-enabled applications and systems. We need even more smart people like them.

There was evidence everywhere that the firms in the OpenStack community are hiring, especially for software engineers. As innovation occurs and new job markets emerge, the supply of skilled talent must shift to meet the need. This is a fundamental principle of our global economy and is clearly on display in high-tech, specifically the world of OpenStack and cloud.

What is my call to action for software developers interested in OpenStack?

I’ll close with a call to the software development community: if you have an interest in developing the latest, emerging cloud technology, OpenStack represents a terrific opportunity to contribute to a meaningful project that has the potential for long-term global impact. There are lots of ways to contribute. In fact, Internap is doubling our OpenStack-related engineering team. Who will step up to build OpenStack into a leading cloud platform of the future?

What makes the iPad® and Kindle Fire different?

I continue to wonder at the beauty of the iPad®. It’s slick and powerful and comes with a price tag to match. Newly announced Amazon’s Kindle Fire is not as powerful or slick, but can access content just as well as the iPad can for just $199! The most amazing part about the Kindle, at least to me as a product manager, is its new Silk browser, which is partially to thank for the lower hardware cost needed to run apps, stream content and shop the web. One of the neat things about Silk is its “split browser architecture,” which dynamically determines whether to use the onboard Kindle hardware to process transactions such as accessing a web site, or push them to the cloud where far off computers collectively do the same thing. The end result is that web sites load faster with the combined power of the Kindle and the remotely located computing and storage systems.

Why is speed so important to Amazon?

Increasing the speed of every CIO, whose charge is to run a corporate website or web-based application, has on his or her mind. Amazon is painfully aware of the impact of latency since their testing revealed that every 100 ms increase in load time of Amazon.com decreased sales by 1% and this is a big concern since Kindle is a gateway to Amazon.com.

How is Internap solving the same challenge?

Analogous to the way Amazon tackled these concerns on the tablet, the company I work for, Internap, has tackled the same concerns at the network level. Our TCP Acceleration as a Service (AaaS) not only allows users to load web applications up to 400% faster than without it, but also maintains consistent, accelerated performance throughout their entire session.

Want to share your thoughts?

Want to weigh in? Just comment below…