Author Archive: Joshua Rushe

January 11, 2008

I Need a Whataburger!!

Somebody...Anybody...I need a Whataburger!!

If you haven't been to a Whataburger, I'm sorry. It's an amazing fast food chain that sells not only the freshest made-to-order burgers, but they're also open 24-hours a day, and their breakfast is second to none (Chris Menard has a clinical addiction to their taquitos). The problem with this is that they only exist in the South. I'm in the North. In Seattle, Washington to be precise—accompanied by our go-live team to manage our newest datacenter and make sure the launch goes smoothly.

On the bright side (no pun intended, it hasn't stopped raining since we landed), it has. We have assembled an amazing team, the datacenter is absolutely spectacular, and the locals have been very friendly. Efficiencies we have built into our normal daily operations over the last two years have basically allowed us to "drag and drop" our datacenters as needed, where they are needed without having to reinvent the wheel every time we launch. Since the deployment is simple, we can focus on service upgrades—like the latest 40-Gigabit rack-level connections—while we roll out a new facility. Connectivity you could use…say…to look for a Whataburger near you http://www.whataburger.com/one_near_you.php (I look every day). We've already flown through our first historic Seattle Truck Day, and had a second one to boot. We're provisioning droves of machines for new and current customers who are taking advantage of our network architecture, tools, and StorageLayer to create their own custom solutions. In a nutshell, we have brought a new DC online and maintained the ability to provide our customers with the same cutting edge hardware and innovative utilities that they have come to expect in Dallas.

On the darker side, with everything is going so well, it leaves a lot of time to sit and think about a tasty Whataburger. With jalapenos. And bacon. Ugh.

-Joshua

October 19, 2007

A Well Designed Infrastructure Makes Everyone Green

As we all know there is an incredible amount of attention being paid to the “greening” of IT. Most people in the hosting industry regard this as the responsibility of the datacenter, as they can make the largest impact with their large-scale deployments of energy-efficient power supplies and processors, efficient physical layouts, cooling practices, and recycling.

Outside of the hosting industry the options become more varied—namely the ability to save massive amounts of power by turning off unneeded infrastructure during non-peak times. A great example would be a call center that operates 9-5 and shuts their workstations down when not in use, or an accounting firm that turns off their billing servers when they go home for the day. This is far from a common practice currently, but it is a very logical and easy step to conserving power. The gotcha here is that unless you can physically walk over to the infrastructure and power it back on, you are going to have to call someone to do it for you. Then wait for them to do it for you. Then hope that they don't forget. This leaves many businesses with infrastructure in an outsourced datacenter throwing their hands in the air, because it's frankly just too risky to not have their resources available at 9:00am when their day starts—might as well just leave everything on.

The story is a little different here at SoftLayer. Using our innovative network design and remote power control, our customers are redefining the way that IT is deployed in an outsourced datacenter. They run their web and mail servers here, pretty normal stuff. But utilizing the SSL to private backend network feature (allowing them to completely disable connectivity to the public network), they are also deploying their domain controllers here. And their office file servers. And their central servers to which their local thin clients connect. They are getting them out of the closet in the back of the office and into a datacenter on enterprise-grade hardware. And you know what they do at the end of the day? They turn them off. The next morning, a click on the power control in the SoftLayer Portal brings them instantly back online anytime, day or night. No phone call to support needed, no waiting for someone else to do it for you. The impact of technology designed to give you optimal control of your IT environment is staggering, especially when you see so many companies utilizing it.

So not only can you choose to deploy your operations in a datacenter that is making enormous strides in green infrastructure, but you can also deploy in one that provides you with the ability to control your own impact as well.

And just like that, everyone gets to be green. And sorry, envy doesn't count.

-Joshua

June 26, 2007

TTL and propagation

Every DNS record is equipped with a TTL. The TTL (Time To Live) is basically the expiration date on that record. Long story short, it's a countdown from when it was initially received until when it is marked as invalid and discarded for a replacement record. This is a very important piece of information that I've run into often as being either outright ignored or misunderstood.

Let's say you have a domain-something awesome like awesomedomain.wow--and awesomedomain.wow has a TTL of 24 hours. When I go to visit awesomedomain.wow as a new visitor (and you know I would, because it sounds awesome) I'm going to receive a record translating awesomedomain.wow to an IP address that will be valid for 24 hours. Any other time I visit that domain in the next 24 hours, I'm going to use that cached address because the record hasn't expired yet. In 24 hours regardless of if awesomedomain.wow has moved IPs, I'm going to trash that old DNS record I've cached locally and go look it up again. The new record will then be referred to by me for the next 24 hours, at which time I'll do it all over again.

But what happens when you have to change your IP, but you want your visitors to see the smallest amount of downtime possible? My first suggestion is to mirror your sites on both IPs, but that is a different discussion entirely. The second is to manipulate your TTL. First lower it to something smaller-from a day to an hour perhaps. Then give that new record with that new TTL at least 24 hours to propagate. Now you can be certain that at the 25th hour, all of your visitors now have a record that will expire in one hour. Next, change your IP for awesomedomain.wow, the record that your visitors have cached locally will expire in an hour, and then they will have your new record with your new IP. Feel free to bump your TTL back up to what it was originally in this step, since they have the new IP. Now your visitors have only had an old record for an hour rather than 24, and they probably missed that hour it was inaccessible while they were posing for a painting or having their top hats heightened. Because all of your visitors are terribly classy.

-Joshua

June 20, 2007

An Interview With an Elevator

SL: Good morning, thank you for taking the time to meet with me.
Elevator: Ding.

SL: Excellent. How would you describe the costs maintaining efficiencies in a hosting environment?
Elevator: Going up.

SL: Well, I think that’s obvious, depending on where you start. Perhaps a better way to phrase this would be, “How would you recommend leveraging existing technologies to implement an efficient execution of a hosting environment?”
Elevator: Ground floor

SL: Well said. I agree that it becomes difficult to put solutions into place after-the-fact, and that in order to run smoothly one must start with a solid plan and avoid retrofitting later. That ends up being far too costly and stifles resources a company should be using to grow their product. How would you describe the attitude of most large hosts with regards to “going green”?
Elevator: Please step away from the door.

SL: I too think that many datacenters out there are concerned with “stepping through” as it were to move operations in that direction. But, since the datacenters can hugely benefit from cost-savings due to reduced expenditures for cooling and power, it is very much worth the shift. What factors outside of the DC could play into making this shift easier?
Elevator: Lobby

SL: Well, I’m not sure that lobbying is the answer, though it may help. Really I was asking about computer manufacturers making the shift to properly-matched and high efficiency power supplies and processors. New technologies are making it easier for younger companies to go green, and older hosts are left trying to figure out how they can turn thousands of antiquated servers into efficient appliances. This goes back to your earlier comment regarding starting out with a solid plan making it easier to
Elevator: Ding

SL: Don’t interrupt me. Easier to maintain a plan than adjust and retrofit to a new one.
Elevator: Second Floor

SL: I’m not sure why you said that, it doesn’t make any sense. Having a host that doesn’t play catch-up constantly benefits the customer in several
Elevator: Ding

SL: Stop it.
Elevator: Third Floor

SL: You’re an idiot. I’m going to go interview the printer.

Categories: 
Subscribe to Author Archive: %