Site With The Lamp

Dell is in a lot of hot water for shipping lousy hardware. One wonders why the same fate has not befallen Apple, the company that is so arrogant as to elegantly package a brick and call it an iphone or an ipad and charge you heaps of money for it. Macbooks are only slightly better. It was only a couple of months ago that my Macbook died off. Fortunately I was able to bring it back to the land of the living thanks to Ebay. The price that the local Apple dealers wanted for fixing it was enough to buy a netbook. And the 13″ macbook is only slightly more powerfull (or bigger than a netbook anyway).

Motherboard or logic board or whatever

Yesterday, my father’s mac refused to power up. He had shut it down normally the previous day. I never shut down mine for fear that it wouldn’t start up again and in his case it didn’t. So he tried all the usually remedies like resetting the SMC etc etc but when none of them worked, I was asked to take a look. I immediately suspected a broken cable like happened with mine. After taking the damned thing apart (there are about 20 screws to undo), I found the cable was ok. Anyway to be safe, I tested with another cable only to find that the computer doesn’t start up.

I bet you are itching to ask if I tried a different battery or a different power supply. You bet I did. So the only logical conclusion now is that the mother board (Illogically named the Logic board by Apple is toasted). And guess what? this is the second time that the mother board on this computer has got toasted! The first time around it was still under warranty but that is no longer the case.

There are plenty of logic boards for sale on Ebay. But they are of no use whatsoever.  That’s because Apple being as stupid as they are, have thought fit to solder the CPU onto the mother board. Yes you heard that right, in Apple notebooks you cannot upgrade the CPU unless you are an electronics engineer. Even then most electronics engineers may not have the equipment needed to replace the CPU or the motherboard at home.

The best thing to do with this macbook seems to be to cannibalize it for parts, any of the macbooks that belong to other family members or close friends are sure to develop some fault before long.

CDNs exists because it’s widely believed that they help speed up website load times by distributing static content to locations that are geographically closer to the user. However this might actually slow things down for some users. Typically you can use tools such as Firebug, Chrome Developer Tools and Yslow to check your page load time and efficiency. But once again you are testing from your own location. Fortunately there are a few tools that will help you figure out the page load times from remote locations. They include Pingdom Tools, Show Slow and WebPageTest. WebPageTest even allows you to specify different locations for the test server. However the best looking tool appears to be Boomerang. But it’s still early days for Boomerang so let’s look at a few results from Pingdom Tools.

Pingdom test while cloud front is enabled.

Static content hosted on S3

Without and content distribution

All the images can be clicked on to be zoomed. As you can see, content distribution really does have a positive impact when the site’s visitors are from the United States.

CloudFront, built upon Amazon S3 is a useful cloud hosting solution because there are no monthly commitments. You pay for the bandwidth you use. Well you do end up paying twice for bandwidth unless you are smart enough to use proper cache control mechanisms. Even then CloudFront is probably more suitable for small businesses than other Content Distribtion Networks like Akamai or Cachefly because they all have monthly commitments.

It is debatable whether using a CDN will speed up your site. In fact, I just discovered that using Cloudfront actually appears to be slower than just using S3 or indeed delivering static content from an Elastic Compute Cloud server.

Page load times all content on an EC2 server

Static content from Amazon Cloud Front

http://raditha.com/blog/images/aws/s3hosting1.png

The reason for this apparent slowdown is that CDNs are generally geared to send you route you to the closest geographic location. The closest geographic location may not always be the closest network path. Even they are the same, it’s still possible that there is less bandwidth in the fiber optic cables for that location, as a result if lots of surfers are connecting to that location, you might find things a little bit slow.

Users from Sri Lanka are directed to the CloudFront servers in Singapore. Even though Singapore is much closer to me than the United States, most traffic from Sri Lanka ends up heading Westwards towards US than eastwards towards Singapore. The same applies for most parts of South Asia including India.  Most ISPs in different parts of the world are better equipped to route traffic towards the US than towards other countries for the simple reason that historically a vast majority of web servers were located in the US. In fact the business model of small web hosting companies is to rent out racks full of servers from large us hosting companies and to sublet them to their customers. Which means, whenever you connect to the website of the office across the street from yours, you are actually accessing a server in the United States.

When twitter’s @anywhere came out many geeks (yours truely included) thought of ways of adding it to their sites and blogs. @anywhere has a lot of usefull features, then again some it isn’t very usefull either (unless of course you are building a fully featured twitter app (which most website owners aren’t doing)). In the end like many others I ended up using it to create a ‘follow me on twitter’ button. Sure it’s a bit more dynamic than a static image but it’s a lot bigger than a static image as well.  There is a difference of about 20Kb when the page loads with and without the @anywhare javascript.

Page size with the @anywhere javascript included

Page load statistics without the @anywhere code

Sounds bad huh? well maybe, but not even half as bad as Facebook’s like button. It adds about 70Kb to the page size and doesn’t do anything other than display the like button. In the case of @anywhere, it’s got a load of features it’s just that we are not making use of them.