Comments for Application Performance Engineering Blog - Shunra Software

Comment on Running mstest without Visual Studio. by Ashiq

Ashiq — Tue, 07 Feb 2012 19:28:19 +0000

Amazing post, thanks! I was getting fed up of discovering and registering all those DLLs manually; you should maybe provide them in a ZIP as an “MSTest stand-alone.”

Comment on Mobile performance engineering rules for the iPhone by Unleashed Technologies uses Shunra’s Applications to Optimize Mobile Site Performance | Application Performance Engineering Blog - Shunra Software

Mon, 30 Jan 2012 17:40:17 +0000

[...] Additional best practices can be found in Shunra’s blog Mobile performance engineering rules for the iPhone. [...]

Comment on Being slow is not hip! by Performance Calendar » Secrets of High Performance Native Mobile Applications

Sun, 11 Dec 2011 07:11:18 +0000

[...] applications that exhibit similar performance faux-pas. Hipmunk, the hip flight search application, downloaded a big data file (650KB after compression), containing the entire search results in one chunk. It would have been [...]

Comment on Etsy’s Mobile App – Attack of the Clones by Gene Litt

Gene Litt — Fri, 25 Nov 2011 21:49:43 +0000

I hope Chad Dickerson is paying attention – Etsy has done a wonderful job trailblazig Continuous Deployment, but they’re overlooking Performance Engineering. PE can be automated prior to pushing the button – performance can be designed in. Or…just keep pushing the button and continuing to put poorly performing code into production.

Comment on APM is Broken by James

James — Thu, 03 Nov 2011 13:49:10 +0000

As a performance engineer and person on a team to bring APM to our company I do agree that there is a disconnect between the sales pitch and delivered product. But, that is always the case with most products; look at the pictures of food on a menu.

I tend to disagree that APM is broken; the processes, the teams and unrealistic expectations, even the sales rep. might be. We are going through a major APM implementation and are seeing plenty of benefits already. That being said it is being implemented by a team of highly technical Architects and Sr. Analysts.

APM and APE should go hand in hand and they are not mutually exclusive. If you success is driven by your software it is foolish to ignore either. Our plan is to have APM on our QA environments as well as or production environments.

Comment on Running mstest without Visual Studio. by Mark Kharitonov

Mark Kharitonov — Thu, 20 Oct 2011 20:53:50 +0000

Hi Deepu.

We no longer use mstest as our unit test engine. We have long since ported all our unit tests to the excellent MbUnit engine as provided by the Gallio platform (http://www.gallio.org/) I highly recommend it.

I am sorry for being unable to help you with your question.

Comment on Running mstest without Visual Studio. by Deepu

Deepu — Thu, 20 Oct 2011 09:30:50 +0000

Hi,

I am working on unit test automation using MSTEST, i would require a help to export the MSTEST execution results to Excel and SQL, if you guys have any approach to this using any code, please share it with me….

Thanks you,
Deepu

Comment on Fitfu for iPhone – Bigger, Stronger but not Faster by Israel

Israel — Thu, 15 Sep 2011 14:42:16 +0000

Thanks very much, I appreciate your thorough response – don’t hesitate to email me if you have any other questions. I look forward to using future versions of fitfu!

Comment on Fitfu for iPhone – Bigger, Stronger but not Faster by Benjie

Benjie — Thu, 15 Sep 2011 13:56:54 +0000

Great analysis of our networking, though our claimed speed improvements were in the app’s interface and exercise processing code so as to enable it to run on older iOS devices (a frequent request) rather than to the networking speed.

The large uploads, as you note, are compressed – we don’t bother the server/client with compressing small requests as the payoff is not significant enough. (We have a defined size such that if the request exceeds it then compression is automatically enabled – IIRC this is around 1KB.) The app does real time analysis of the accelerometer data (the counting aloud) but this data is uploaded in order to allow deeper analysis of the motion in future and also to help us to hone our motion engine (also to address accusations of cheating!).

I had started working on ‘request rollup’ functionality into the API client late last year which would gather all parallel requests sent over a small time period (say 0.5 seconds), bundle them into a single HTTP request and then split up the aggregate response so that the process would be completely transparent to the rest of the application’s code. However I’ve been suffering from a long term illness, so work on this was put on the back-burner and we’ve been concentrating on the interface and features instead. I hope to return to it in future (we are only a small team!).

The avatar is downloaded and cached (the cache was faulty prior to 1.7) so that you can see the full avatar when you click on it in a user’s profile (one file download instead of many at different resolutions – I’m sure you’d approve of this!). In future we plan to fully outsource this avatar caching (using ETag headers and the like) to AWS CF/S3 – the current solution is an interrim one to maintain compatibility with existing versions of the app whilst we work on the new code.

We’ll definitely look at removing some of the analytics calls though!

By the way, we ensure that only one large upload happens at a time to prevent network congestion – this keeps the apps networking responsive whilst still getting the heavier data to us. We did consider only uploading the larger files when connected via WiFi, but it seems some people are very infrequently connected to WiFi and so we decided to avoid this so as to prevent network traffic burts.

Comment on APM is Broken by marty.brandwin

marty.brandwin — Wed, 14 Sep 2011 17:40:48 +0000

Joel,
I agree with your comment. When we say “APM is broken”, we are using Gartner’s definition of APM which is “application performance monitoring” – the tools or solutions used in the production environment. You are correct, though – the “application performance management” process is also broken. The feedback loop you described is not often fully or correctly implemented: the disconnect between pre-production and production environments (and people) must be fixed. For example, pre-production test labs must be able to incorporate real-world production network conditions in order to ensure reliable testing (load, performance, capacity, etc.). Similarly, operations teams that are monitoring applications in production must know the SLAs that have been established and validated in pre-production so they know what a performance error is.

Thanks for the comment and we look forward to continuing the dialog.