Tuesday, July 29, 2008

How to Recover the Missing X(1) for the USL Scalability Model

When it comes to assessing application scalability, controlled measurements of the type that can be obtained with tools like Grinder or LoadRunner, are very useful because they provide a direct measurement of the throughput, X(N), as a function of the vuser/generator load, N. These data can be input easily into my universal scalability model (USL). To apply USL, however, you need to normalize all the X(N) data to the X(1) value. It often happens that the X(1) value may not have been measured by your QA group or it simply may not be measurable easily. What do you do in that case? You estimate it (carefully).


Friday, July 18, 2008

Mr. Amdahl Calls the Repairman

Back in January, I reported that my Universal Scalability Law (USL) had been proven to be equivalent to the synchronous throughput of the load-dependent machine repairman queueing model. Although I wasn't lying, what I really had was what we call a "sketch" for a proof. The details needed to be filled in, but that was just a question of finding the time to do it. In May, I found time to write the first draft and thought I would be finished shortly thereafter. That attempt ground to a halt for two reasons (you might want to refer to Appendix A in my GCaP book for the background):

  1. Simulations that supported the theorem involved a barrier synchronizer, and for this to work continuously, my colleague Jim Holtman, discovered that the distribution of parallel execution times (with mean 'think' time Z) could only be deterministic (D). That's a special case of the repairman: D/M/1//N, whereas the equations in my proof show quite clearly that it must work for M/M/1//N.
  2. There's a theorem in queueing theory (Bunday and Scraton, 1980) which states that what holds for the repairman (M/M/1//N) also holds for a more general source of requests viz., G/M/1//N. Since G is a superset of the distributions D and M, this theorem suggests that our simulations probably shouldn't be restricted to D-distributed think/execution times as observed in (1). Or should they?

Musical Chairs at AMD

It looks like Intel's very aggressive pace for technology advancement, which I have been tracking loosely over the past 18 months, is really starting to hurt AMD. The response from AMD, however, may represent continued dithering at the executive management level.

Monday, July 14, 2008

Guerrilla Data Analysis Class - Seats Still Available

Most operating systems are capable of collecting hundreds of system and performance metrics every second. Even if you only record them once an hour, after a week you will have more than 50,000 samples; after a month you will be staring at almost a quarter of a million samples! But data are just data. How do you extract performance information out of all those data? Easy! You attend our new and expanded 5-day Guerrilla Data Analysis class, here in California.

In this class, computer engineering and statistics expert Prof. David Lilja presents an easy introduction to statistical methods and finally leads us into the topic of Design of Experiment (DOE) methods applied to performance and capacity planning data.

Having established the foundation theory, R expert, Jim Holtman will show you how to apply DOE and other statistical techniques using example case studies.

You can register for the class, and book your hotel room, online. Book early, book often! We look forward to seeing you in August.

Saturday, July 12, 2008

Apple IT Dies on iPhone 3G Launch

It's great to be popular, but hell to be the rage.

In the aftermath, Apple managed to sell 1 million iPhones in just 3 days cf. the same number of the original iPhone in 2.5 months. Why should you care? Mobile devices like this are actually computers, not just phones. The iPhone runs MacOS X; same as mac laptops. In my view, this and similar devices represent the commodity computer of the future.

Wednesday, July 9, 2008

Back from the Old DDR

Read this to the tune of "Back in the USSR".

I spent part of last month in eastern Germany (former DDR) where there is a decided lack of internet connectivity (hence, no blog entries), even after west Germany (FDR) has thrown a considerable amount of euros to bring the eastern economy up to speed. Hotels either don't offer internet at all or want an arm and a leg for it, if they do.

As far as I can tell, eastern hotels (most of which really don't have the inclination to understand IT; just like hotels in the USA) have let Deutsche Telecom swoop in monopolitically to set up internet packages whereby they can charge 9 euros (a whopping $14 USD) for just 1 hour online, unless you are a card-carrying DT customer, in which case you are entitled to some kind of discount. After resisting for days, I eventually broke down just to prevent my web mailbox from overflowing. These packages are affectionately termed "Hot Spots"; a term possibly meant to confuse tourists like me who naturally jump to the conclusion that it means New York style free wi-fi. It's wi-fi ("WLan") all right, but at a very hot price.

Amazon and Google Discover Erlang

Erlang the language, that is.

VMware CEO Walks

VMware CEO, Diane Greene, stepped down abruptly on Tuesday, sending shares down nearly 25%. No reason was given but the company had warned revenue for the quarter ending June 30 would be below its predicted 50% growth compared with last year. It could also be partly a reaction to the commercial marketplace heating up with Microsoft, Sun Microsystems, Hewlett-Packard and Citrix (backed by IBM) having all built or acquired similar virtualization technologies.

On the FOSS scene, we have VirtualBox as a replacement for VMWare Fusion or Parallels.