GigaOM » web site performance

Yottaa Uses the Cloud to Boost Site Speed

Derrick Harris — Wed, 27 Apr 2011 14:26:12 +0000

Updated. Yottaa, a Cambridge-Mass.-based, web-performance startup, announced its Yottaa Optimizer service Wednesday, which promises to improve website performance by leveraging a global network of cloud servers. Yotta claims the service delivers results “with just a few clicks,” which would make it a welcome alternative to many website operators without the knowledge or resources to utilize content delivery networks such as Akamai or Limelight Networks.

That being said, the company promises it can cut page-load times in half, even for customers already using CDNs or load balancers. It does this by intelligently routing jobs across its “global elastic multi-cloud network … using more than 100 techniques such as domain sharding, CSS spriting and content delivery network (CDN) balancing.”

I have reached out to Yottaa to confirm its cloud providers and locations and will update this post when I hear back.

Update: Yottaa uses 11 different cloud providers for Yottaa Optimizer, including Amazon Web Services. During the recent outage, a company spokesperson told me, Yottaa was able to reroute AWS operations to other clouds to avoid any lengthy downtime. It plans to automate this capability for customers in the future.

The company also uses AWS to host its Yottaa Insight service, which tests and monitors website performance and provides suggestions to improve load speed. That product won Yottaa a spot as a finalist in the 2010 Amazon Web Services Start-Up Challenge.

Yottaa’s service highlights one of the big sweet spots for cloud computing, which is the ability for startups to compete with CDNs and other service providers without having to build their own infrastructure. By writing smart software and leveraging cloud providers’ already-built infrastructure, companies such as Yottaa can keep their overhead down, which results a lower cost service for customers, too.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

Twitter Moves Into Data Center. Goodbye, Fail Whale?

Derrick Harris — Mon, 21 Mar 2011 20:26:37 +0000

Making good on its pledge last summer, Twitter has moved into its own custom-built data center designed to handle its unique needs, and which the company claims will be the service’s “final nesting ground.” Twitter engineer Michael Abbott detailed the move in a blog post this morning. With the process complete, Twitter has few excuses left for extended or frequent downtime, something for which the popular service has become known over the years.

The migration, as Abbott describes it, began in September and required a fairly detailed process of replication and careful staging:

First, our engineers extended many of Twitter’s core systems to replicate Tweets to multiple data centers. Simultaneously, our operations engineers divided into new teams and built new processes and software to allow us to qualify, burn-in, deploy, tear-down and monitor the thousands of servers, routers, and switches that are required to build out and operate Twitter. With hardware at a second data center in place, we moved some of our non-runtime systems there – giving us headroom to stay ahead of tweet growth. This second data center also served as a staging laboratory for our replication and migration strategies.

At the same time, Abbot explained, Twitter’s engineers were busy prepping the ultimate data center to take over all of the work:

Next, we set out rewiring the rocket mid-flight by writing Tweets to both our primary data center and the second data center. Once we proved our replication strategy worked, we built out the full Twitter stack, and copied all 20TB of Tweets, from @jack’s first to @honeybadger’s latest Tweet to the second data center. Once all the data was in place we began serving live traffic from the second data center for end-to-end testing and to continue to shed load from our primary data center. Confident that our strategy for replicating Twitter was solid, we moved on to the final leg of the migration, building out and moving all of Twitter from the first and second data centers to the final nesting grounds.

As Abbott explained, the new Twitter infrastructure — which reportedly is housed within C7′s new Salt Lake City data center — is designed to optimize performance and give Twitter extra runway for improvements. For Twitter’s sake, the strategy better pay off because all of the new features in the world won’t necessarily make up for relatively regular bouts of downtime, especially as the service grows even more popular. If the migration process is any indication, Twitter may finally have found the cure for the all-too-common fail whale, as Abbott cites improvements in performance and uptime during the migration, while still adding new features and more users.

Twitter’s new infrastructure also highlights a big difference between itself and Facebook in terms of sheer operational size. After all, when Facebook started to outgrow its hosted infrastructure, it didn’t lease space within a colocation facility; rather, it built its own cutting-edge facility. What’s more, it doubled the planned size and sited a second facility during the construction process. Facebook has more users, houses much more data and is rolling in cash, so its out-of-nowhere decisions to build two large data centers isn’t too surprising. But I wonder whether its fate doesn’t foretell what might happen with Twitter a few years down the line. Will continued growth, data and performance needs take Twitter to a place where it’s in the company’s best interests, technologically and financially, to break ground on a new final nesting ground?

We have approached Twitter for a comment on this story and to verify the location of its new data center.

Image courtesy of Flickr user TheMuuj.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.

Tracking Web Performance Can Make or Break Online Ventures. Some Tips To Pick The Right Tool

Yadid Ramot — Sun, 02 Aug 2009 15:00:50 +0000

In our era of immediate gratification, creating a quick and responsive web experience is often as important as the product or service you’re trying to sell. Poor performance can frustrate users to the point that they will leave your site. Recent studies show that on average, 20 percent of a page’s load time is spent on the back end as the server processes and generates HTML, while the remaining 80 percent is spent on the front end while the browser tries to load page assets and render the HTML. This rule applies to any size web site, whether it receives 100 or 1 million visitors a day. Keeping in mind this 20/80 rule, we at MySpace give as much attention to the front end behavior of our pages as our back-end systems.

A lot of tools and services offer front-end performance tracking, and they vary by feature and pricing. Selecting the right tool can be confusing, especially if you’re new to the performance-tracking field. Different technologies and methods are available in today’s market. Here are a few tips you should take note of when choosing the right tool.

In general, three types of front-end problems can result in a slow browsing experience: network/socket bottlenecks — bottlenecks that are caused by a bad network and or download configuration; positioning the page’s elements in the wrong order; and heavy client-side processing — bottlenecks caused by overloading the client’s CPU and memory.

The most common bottlenecks are the first two: network and socket bottlenecks and bad content delivery prioritization. The method for tracking these two problems is to load the site into a browser and capture the download order and network traffic of all the elements on the page. Capturing the network and download order is usually accomplished by installing some type of network sniffer on the client’s machine or by setting up an HTTP proxy between the client’s machine and the web server, then tracking the inbound/outbound traffic. The following is a list of key features you should take note of when you choose a tracking application for your site:

The application must provide a “waterfall graph” of downloaded files. Waterfall graphs are extremely helpful in identifying which download was done in parallel and which was a blocking download.
The application must measure server wait time, DNS lookup time, request send time, and response receive time. It’s important to measure these numbers to get a better view of what needs to be improved.
The tool must track the data using either an actual web browser or a browser simulation. Select a tool that uses the same download logic as your client’s browser. Some browsers are capable of downloading files in parallel from multiple locations, while others limit the number of sockets. So if most of your clients use Internet Explorer, you must select a tool that either simulates IE or tracks the data using that software.
When running the tool, you should use the same network environment as most of your clients (same geographic location, connection speed, etc.) to get an actual view of what your users are experiencing. For example, if most of your visitors come from the UK using a 256kbp ADSL modem, your application must track the data using a 256kbp modem from that region as well.
When you measure your pages, you must test them using both an empty and a primed cache. The general assumption is that 20 percent to 50 percent of your incoming requests are being done with an empty cache. This supposition was proven to be true in a test that Yahoo conducted.

Here are a few tools worth checking out:

HttpWatch (Starting at $395) HttpWatch is an HTTP viewer and debugger that integrates with IE and Firefox to provide seamless HTTP and HTTPS monitoring without leaving the browser window.
Google’s Page Speed (Free) Page Speed is an open-source Firefox/Firebug add-on. Webmasters and developers can use Page Speed to evaluate the performance of their pages and get suggestions on how to improve them.
Firebug (Free) Firebug integrates with Firefox to put a wealth of web development tools at your fingertips while you browse. You can edit, debug and monitor CSS, HTML and JavaScript live in any web page.
AOL’s Pagetest (Free) Pagetest is an open-source tool for measuring and analyzing web page performance right from your browser. AOL developed Pagetest internally to automate load time measurement of its many sites, and it has evolved into a powerful tool for developers and software engineers in testing their web pages and getting instant feedback. The company decided to release it to the grander development community to further help evolve it into an even more useful — and free — web performance tool.
MySpace’s MSFast (Free) MSFast is a browser plugin that helps developers improve their code performance by capturing and measuring possible bottlenecks on their web pages. It is custom-built technology that enables any developer to track page load performance and user experience data. The open-source tool measures the CPU and memory consumption of rendered HTML. It also reviews downloaded files and shows download time estimation on different bandwidths, and measures the time it takes to render each section of the page in different connection speeds. For more details about MSFast, check out http://www.msfast.com.
Fiddler (Free) Fiddler is a Web Debugging Proxy which logs all HTTP(S) traffic between your computer and the Internet. Fiddler allows you to inspect the traffic, set breakpoints, and “fiddle” with incoming or outgoing data. It includes a powerful event-based scripting subsystem and can be extended using any .NET language.
IEInspector’s HTTPAnalyzer (Starting at $149) IEInspector HTTP Analyzer is a handy tool that allows you to monitor, trace, debug and analyze HTTP(S) traffic in real time. It is used by industry-leading companies including Microsoft, Cisco, AOL and Google.

Yadid Ramot is MySpace’s Client-Side Performance Architect. He spends his time mentoring the MySpace development community on front-end performance and developing performance-monitoring tools.

Related research and analysis from GigaOM Pro:
Subscriber content. Sign up for a free trial.