04.01.2006 : XML Infinity Goes Faster

The XML Inifinity server had a speed increase given to it today as a result of a new procesor coming out of testing as well as rationalisation of the code base. The predominant speed increase will be noticed within the Translation Engine but overall the system is performing faster than what it was.

Andrew Fisher, XML Infinity Chief Engineer explains further, "Since the launch of XML Infinity version 3 the guys haven't sat back on their laurels at all. They have gone through all of the code and applied changes, particularly within the Translation Engine.

"The Translation Engine is the slowest process because of the sheer amount of processing that needs to be done to take the base XML and apply the templates to it. It also has to be done potentially hundreds or thousands of times in order to translate an entire site. Thus a small improvement in the base code can lead to huge improvements in full-site translations.

"The changes the guys have made within the code base have accounted for about a 5% increase in speed. Everyone is really pleased"

The second major speed increase has been due to a brand new processor, in the shape of an AMD Athlon X2 4800+ chip. AMD were one of the first to market with a commercially viable, and not prohibitively expensive, dual core processor and the results speak for themselves.

Andrew says, "Back at the start of the migration to .NET and the whole specification for version three of the system we looked at all of the emerging technologies and said 'What is going to work best for our system' - it is rare you get to work from such a clean slate.

"Something the entire development team discussed at length was the new breed of dual core processors - these look identical to a normal computer's CPU but have two chips inside them instead of one. What this means practically is that the processor can execute two things at once.

"You might think this is something your computer does already but it doesn't. At the end of the day 1 processor means one instruction executed per time cycle [usually 1 billionth of a second]. What Windows does is make it appear it is doing multiple things by acting like a traffic cop and sharing this resource around to everything that asks. You then get into a prioritisation war between applications where things get stopped whilst control is handed over to another process for a short period of time.

"A multi-core chip turns all of this on it's head as it can execute things at the same time properly. By using threading [creating new units of execution for difficult tasks] on the translation engine it means we can do more translations at the same time."

Whilst this all sounds very techie and complicated Andrew assures us it isn't. "Think of this as the old adage 'Many hands make light work' and you have exactly what is going on - but you have to program the system to work this way as it won't do it by default".

So in real terms there has been a dramatic improvement. At XML Infinity we have a test site we use to benchmark the translation engine. It consists of over a thousand files and is an incredibly complex set of data and translation templates so any benefit felt on this test set will have real benefit for end users. On the test set the full translation time went down from 82.5 seconds to 36 seconds so the decision to go down this route was fully justified.

Contact: news@xmlinfinity.com

< Back to news archive

User Login

Infinite News