Performance – bish.co.uk

Monitoring Ubuntu Linux from LoadRunner (RSTAT)

Over the years I’ve used RSTAT to monitor the performance of Linux servers during performance tests many times. Historically I’ve asked Linux admins to enable RSTAT for me but over the last year or two I’ve incorporated various Linux machines into my test and demo environment so I’ve had to do this myself.

I’m primarily a Microsoft specialist so I always seem to rely on Google searches and some trial and error to resolve problems that I encounter with Linux and this weekend was no exception. I spotted a FaceBook post from Scott Moore describing problems that he was having with LoadRunner and RSTAT so I decided to see if I could help him.

I installed RSTAT on my Ubuntu test machine at home and immediately saw the dreaded error message “Error while creating the RPC client. Ensure that the machine can be connected and that it runs the rstat daemon (use rpcinfo utility for this verification).”

Whatever I tried, I could’t get RSTAT to start. After some ‘Googling’, I found a number of posts in various places which helped me to resolve this problem.

A post on StackExchange (related to NFS issues) told me that rpcbind may have a dependency on a package called nfs-common. I installed this “just in case”. The same post told me that to start STATD (which I think is related to RSTAT) automatically at boot, I needed to add “NEED_STATD=yes” to the file “/etc/default/nfs-common”.

A post on GadgetWiz.com told me that I should edit my /etc/hosts.allow file to ensure that the local host could make rstatd requests.

After restarting my Ubuntu PC, I checked that RSTATD was running using the commands “rsysinfo localhost” and “rup localhost”. I was pleased to see that after making these changes, it was possible to monitor my Ubuntu machine using LoadRunner.

I repeated my test on a new Ubuntu 16.04 LTS machine which I monitored using LoadRunner 12.53. A list of the commands that I used is below:

Install rstatd and related components:

sudo apt-get install rstatd rstat-client nfs-common

Add a line to /etc/hosts.allow to allow certain hosts to make rstatd requests:

rpc.rstatd: localhost

Added a line to /etc/default/nfs-common to start STATD automatically:

NEED_STATD=yes

These commands confirm that RSTATD is running:

rpcinfo -p localhost 
rsysinfo localhost
rup localhost

TFL’s new website coping well during #tubestrike

I recently read an article in ComputerWeekly describing how TFL had redesigned their website using HTML5 to optimise performance across multiple device types. I was interested to see how the new site was handling the likely increase in traffic due to the tube strike.

Prior to the HTML5 re-write the last major redevelopment of this site had been in 2007, well before the proliferation of mobile devices, now used daily to check for travel updates or plan journeys. 75% of Londoners visit the TFL website regularly and there are 8 million unique visitors per month.

I was in London earlier this week so I, along with millions of other commuters, wanted to keep up to date with the news of the tube strike. I, like many others, turned to my smartphone for answers. The site performed well on my Android phone, so I wondered whether the increase in traffic had caused any performance degradation.

At Trust IV we have developed an in-house application to test the performance of websites, we monitor performance for several hundred sites, each of which is categorised into a relevant business sector. TFL was already being monitored in our “travel” category. I was impressed to see that the website was the 3rd fastest travel site monitored today with a page response time of <1.2 seconds. The site developers should be pleased with themselves.

Although occasional spikes in response times were observed (which is common when monitoring in this way); on the whole the site remained responsive throughout the day. Average response times appear no slower today than they were last week (the chart below shows response times in milliseconds).

If only more of the sites that I visit regularly performed as well as this.

Get in touch for more information about our “Test The Market” monitoring application and how it can give you insights into your own website performance and see how your performance compares with your competitors.

See more articles like this, and download the response time report at:
http://blog.trustiv.co.uk/2014/02/tfls-new-website-coping-well

Evian’s “Baby and Me app” performs like a geriatric…..

According to the Danone website, “Baby & Me, created by BETC, shows a street full of people rediscovering their inner youth – and dance moves – as their reflections reveal the baby versions of their adult selves.” If you haven’t seen the advert, you should watch “Baby & Me” on YouTube, it’s fantastic; a perfect example of a viral marketing campaign with over 59 million YouTube views. This advert builds on the success of their “Roller Babies” campaign from a few years ago (69 million views). I’ve shown it to my children and my wife and everybody I’ve mentioned it to seems to have seen it. Evian is reaching a mass market with relatively little TV advertising spend.

To capitalise on the success of the advert, Evian has commissioned the production of a “Baby and Me” app for iOS and Android. On the face of it, this looks like a great idea. The application takes a picture of you as an adult and then uses pattern matching to identify over 70 facial features in a database of baby pictures. The application then creates a composite image of you as a baby. This has great potential for social sharing of images and it isn’t hard to see how this could help to ensure that even more people get the “Release the inner you” message from Evian.

There is no doubt that viral media is a great marketing tool and mobile apps are a great way to use peer interaction and socialisation to spread a message quickly. All good news so far from a marketing / PR perspective, but…….

…….There’s always a “BUT” isn’t there?

The Evian app is sadly disappointing.
Anybody who downloads an app that doesn’t work first time is highly likely to delete it and never use it again. Negative publicity from colleagues and friends who can’t use the app will do a great deal to damage any positive marketing messages that the app was intended to create

It is highly likely that the application designers and developers have been forced to work to a difficult schedule to get this application to market. Whilst individual teams may have all tested their application components; it is highly likely that proper end-to-end testing, including testing over relatively slow 3G and Wifi networks has not been done. Performance testing is crucial for applications like this, a poor user experience will result in users abandoning the application and never downloading it again.

So what went wrong?
Whilst performance testing is a well-established profession……
…testers need to evolve.

15 years ago testers needed to understand thick clients and complex client-server apps.
10 years ago, web applications dominated the market and testers skills (and pay rates) started to fall.
5 years ago, Web 2.0 and mobile technologies started to develop and many testers failed to understand the complexity of more complex multi-tier web applications.
Now testers need the ability, skills and tools to test across multiple devices and multiple networks and need a deep understanding of application architecture to ensure application stability, scalability and performance under load.

Many testers don’t have the required skills or testing tools to test complex multi-tier applications across different devices and network topologies. Evian have used big-hitting media companies, BETC Digital and B-Reel, to develop their application…. I bet they’re wishing that they had the application tested properly by an independent testing specialist before this week!

The