Designing for cacheabiltiy

One of the things we kept in mind while working on the new site was how make sure that the only content that you have to load when you come to the DevCentral site is the dynamic, new content.  We chose a design that puts the content first, the menu bar is smaller and includes links to the most viewed topics on the site. We made our header much shorter than the previous design. We also narrowed the right hand column, removed the big images.

By moving the content around and into modules, we’ve made it easier to cache different parts of the page. So even though we are showing more content on the page, we can control how long each content section is cached by our CMS. The majority of our design is done using Cascading Style Sheets (CSS) rather than images. Some design elements, like the rounded corners at the top and bottom of the page as well as the containers, is rendered either using CSS or using a JavaScript plugin. By not using very many images in our design, we made it easier for our Web Accelerator module to cache the CSS and JavaScript files. This means that, instead of our servers sending the files to your browser, we let our WA handle serve them directly out of it’s cache. You can see this in action if you view the source for our website in your browser.

<link rel="stylesheet" type="text/css" href="/Portals/_default/Skins/DC5/stylesheets/vader/jquery-ui-1.8.vader.css;pvf98cbd891d9740ac"> 
<link rel="stylesheet" href="/Portals/_default/Skins/DC5/stylesheets/devcentral-release-min.css;pv8fee53457567fdaf" type="text/css" media="screen"> 

Those funky little character sequences at the end of the url to our stylesheets are used by WA to track the version it should serve out. Every so often WA will empty out it’s cache and reload the CSS and JavaScript files from our server. It does this without any of the DevCentral team intervening. Wa also tells your browser how long to cache a file and when to expire your browser cache and re-fetch the file. As you may know, the perceived speed of a web site has almost nothing to do with how much bandwidth is available from your ISP or how fast your connection is, but instead is determined mainly by how many roundtrips your browser has to make to our server.

That being said, how long it takes to download our page DOES matter. Next I’ll talk about how we compress, minimize, and otherwise shrink our page down as much as we can so that we can deliver our content to you faster.

Enable people to connect with other people

One of the things we saw in the forums was people asking questions and people, other than F5 employees, answering the question based on REAL WORLD experience. We wanted to make it easier for people to share their real world experience, both on our site and in the real world. We decided to make the site more social. Not social in the Facebook sense, you can’t play “Router Wars” or grow a WebFarmVille” and poke people on DevCentral. But we wanted you to be able to connect with other people and find out what they are looking at and what they find interesting. We also wanted a place for the regional and local user groups to post information.

Your profile

Your Devcentral profile has been expanded and allows you to share more information with other users if you choose too. Does that mean you need to start posting pictures of your kids? Not unless you want to. But say you have a diagram of your network you want to share, now you can post it to your profile. Have you come up with a new way of keeping your datacenter cool? Post about it in your profile. You can also “microblog”. Did you answer a question on serverfault.com about iControl or ASM? Post a link. You can connect with other users by adding them as a friend. Whenever one of your friends posts something you’ll get notified. You can also send messages to each other through the system.

Topic Groups

The idea behind topic groups is to have a common place for people to talk about, and post about, common themes. So questions about iRules, links to good information about  iRules, and reference. We wanted them to be one-stop shops for information about certain things as well as a way for people to show their common interest. Now you can connect a share information in places other than the forum. Did you write a great blog post about an iRule you wrote? Post a link to it in the group.

User Groups

Sometimes it helps to talk face to face with people about problems you are having or to share a cool new idea you have for a configuration. User groups are also great places to find people if your employer is hiring. By providing a place for regional and local user groups to post information, we giving you a leg up to form your own local user group and letting you know if their is a user group in your area.

It’s about you

The social aspect of the site is the driving point of this release. It truly is all about you. We’ve created several tutorial videos showing how you can contribute. We wanted to create a highly visible area where you can share your knowledge with the world. So link to your technical blog. Post in the forums. Post status updates. Add people as friends. Comment on activity in your stream or in others streams.

Make the navigation simpler

We knew that people had a hard time finding information on the site because sometimes WE had a hard time finding information on the site. We talked with our MVPs and looked at where people were spending the most time at on our site. The number one destinations on our site were the forums, the tech tip articles, and the wiki in general. It was pretty easy to get to the forum you wanted to get to, just a couple of clicks. But sometimes you had to hunt for a link to the Wiki and it wasn’t always easy to find tech tips on specific topics.

The whole team met up in Seattle to discuss the new site organization. We decided to go with a two-tier navigation at the top. We also decided what would be on each level. We knew we wanted to have a section where we could put the latest news and collections of links to various other parts of the site about topics that were hot. We put a lot of thought and effort into what would be on the menu when you first arrive at the site. We knew we wanted the tech tips, forums, and blogs on there as well as a link to the wiki. We came up with the idea of a “Getting started” section to help out people who were new to DevCentral.

We also knew that we wanted a place to put useful links, to your profile, your groups and other useful places. We put this at the bottom of every page.  Over the course of time, you should look for some changes in this bar. We may swap in new quicklinks based on demand. We might also put special notifications in this bar. We’ve got lots of plans.

We also changed the search so that you didn’t have to leave the page you were on to see your search results. There are some issues with the search results that Jason addresses in this post.

By creating topic groups, we’ve made it easier for you to find new and old information about things that interest you. More about groups later.

What do you think about the changes we’ve made to navigation?

Redesigning DevCentral - the general idea

When we sat down to redesign DevCentral we had a couple of themes in mind

1. Make the navigation simpler.
2. Enable people to connect with other people.

I also had a few goals in mind while working on the new skin

1. Design for cacheability
2. The page should load as fast as possible
3. The page should be as small as possible
4. Should be able to extend the site easily

Over the next few days I’m going to write a series of posts explaining how the DevCentral team and I went about achieving those goals.

How LTM VE can help integrate your entire product team.

Right now there is a huge chasm in your product team between your application developers and your network admins. The application developers (appdevs) spend a lot of time and effort building up an application. Then, once they are satisfied that it's ready to launch, they begin talking with the network admins about the best way to install their application, how to upgrade it, how to scale it once it's a big hit. While the appdevs are crafting the application, the network admins are busy building servers, installing operating systems, preparing their "house" for a new baby that they hope won't be too cranky and will sleep through the night right from the start. The two teams are often very disconnected, one hands off to the other.

The network admins work on very expensive, specialized equipment that allows them unlimited control over the traffic that flows through the network. They can use iRules and policies to effectively manage their traffic. Sometimes these rules inject client scripts or html into web pages. Most of the time the appdevs aren't even aware of the advantages that network side scripting can bring to their applications. If you've ever had to write a mod_rewrite rule and make sure that the change to it and synchronized to every server in your farm, you can appreciate being able to write a single iRule that will do the redirection for you. Up until now, only the server admins had access to the BIG-IP's in their organization and they were responsible for writing and maintaining the iRules, even if those iRules were application specific. Now with LTM VE, the appdevs can see how taking advantage of iRules can change the way they architect their applications. Appdevs can run a copy in their development environment with the production iRules running and save some time during deployment. They can work with the system admins to virtualize access to different services they are consuming in their applications.

On past projects, I've had to maintain several configuration files that basically just told the application where to look for certain servers or services. One set for development, another for testing, one more for load testing, one for staging, and finally the production section. Some times each section would have multiple sub-sections in case we needed to work with different combinations of servers and services. For example, you may want to consume a production service, but store the data in a development database to reproduce a bug. You may need to load a customers data set and test your changes against production data to test the bug fix. Being able to work with the network admins to set up app pools in a LTM VE means not having to juggle multiple configurations. You can point at a single location and let BIG-IP LTM VE do all the work.

Joe has a great article about setting up your laptop to do development with BIG-IP LTM VE.

Why should I care about Big-IP 10.1?

Big-IP 10.1 has been released announced. It includes a lot of new features and enhancements to some existing features.

You can read some of the press releases on our main news page:

F5 Delivers Advanced Web Security Solutions to Help Customers Efficiently Address Threats to Web Applications and Enhance Business Performance

F5's BIG-IP Solution Enables Service Providers to Transparently Scale and Grow Mobile Services

F5 Enhances IP Geolocation Capability through Partnership with Quova

F5 Solutions Optimize Microsoft Windows 7 and Windows Server 2008 R2 Deployments

Enhanced support for Windows 7 and Windows Server 2008 R2.

Direct Access and BranchCache

Microsoft DirectAccess lets Windows 7 clients create a secure connection directly to the network over IPsec rather than having to create a separate VPN connection. That means that you can still access your files securely while you are enjoying coffee and a scone at Starbuck Joe wrote a great article addressing how Big-IP 10.1 helps enable DirectAccess using Windows 7 clients with Windows Server 2008 R2 servers in your enterprise. BranchCache is a new feature in Windows 7 and Windows Server 2008 R2 that will cache content from remote servers on local servers. You can access the file once it is cached as if you were working in the remote office, using less WAN bandwidth and wasting less of your time. BranchCache can work in one of two modes: Hosted cache mode and Distributed Cache mode. In hosted mode, a Windows Server 2008 R2 server will host the cached content. In Distributed mode the clients will host the cached content. Big-IP v10.1 will route the file requests to the correct server or client.

Security

DNSSEC

Lori wrote a great post about what DNSSEC is and how it works, “It’s DNSSEC Not DNSSUX”, and Jason posted an article explaining how to configure the new DNS features using both the GUI and TMSH. Configuring GTM Version 10.1's DNS Security Extensions

Advanced bot and scanner protection

Big-IP 10.1 includes support for detecting web site scraping and allows you to set thresholds for locking out IP addresses.

Enhanced reporting

ASM’s new Attack Expert System helps make sense of the attacks you see logged against your network. Each attack has a list of possible attack types and the types are explained. The reporting GUI has been reworked. You can either view a chart or drill down into the attacks with a few clicks. Ten-Point-One also includes a PCI compliance report.

Look for some more information about some of these features in the coming months.

(edited for clarity SCK)

Why should I care about the hardware?

I'm not a network engineer. I just build the applications, the network professionals take care of deploying my applications and making sure they stay up. Why should I care about what hardware they use?

Because the hardware they use could influence my application architecture.

• If I know the IT guys are able to bring in new servers in response to increased demand, I start to think about distributed caching instead of replication maybe. I start to think that the application shouldn't store any data in a machine specific session so that the load balancer can send the user to any machine.
• If they have access to distributed hardware caching, maybe I don't have to worry about donut caching in my pages.
• If I know they can pull servers in and out of the clusters at will with no visible downtime to our users, I start to think about deploying more often. I can become more agile.
• If I don’t have to worry about how much space is left where I’m storing my users uploads, I don’t have to write code to check how much space is available and present an error to my users if the space fills up.

One of my goals with my F5 Devcentral blog is to explore ways that developers and IT professionals can cooperate more and blur the lines between them a bit more.