Performance

This article shows a simple solution to a problem that doesn’t seem to be adequately documented: VS Code UI lags, freezes, and delays caused by the GitHub Copilot extension. Problem: VS Code UI Becoming Slow Having worked on a Python project for a while, I noticed that VS Code’s UI was frequently freezing for seconds at a time. This was happening in various places of the UI: the editor itself but also in the GitHub Copilot Chat window. Copilot also seemed to be taking more and more time getting ready to answer, and the extension-host process would fully saturate one CPU core for long periods of time.

There is more than one way to test the performance of web browsers like Chrome, Firefox, or IE, but regardless of how you do it, you need a consistent workload that makes the browsers comparable. Unless you are testing with synthetic benchmarks (which come with a plethora of problems of their own) you need a way to automate browsers opening tabs and loading URLs. This article presents a simple solution to do just that.

When sizing a new environment it is tempting to use averages. It seems the logical thing to do. But it also guarantees a bad user experience. Example: Sizing an RDS or XenApp Farm Let’s say you’re tasked with building a new Citrix XenApp farm. Being a diligent IT person you set up a pilot: one or two machines with all the right software and settings. Then you carefully select a group of pilot users in such a way that they represent the organization’s employee types statistically correctly. Then you let them work on the new platform, ironing out bugs and such. At the end of that period, you have a great new platform. But there is one big question left: how many servers to buy?!

FileIOTest is a command line tool that tests the speed of local or remote (SMB) storage by performing some common file IO operations repeatedly and measuring the duration. These are the main facts:

In version 2.1 uberAgent collects detailed ICA, PCoIP, and RDP remoting client properties. Yes, that is right, we have added support for VMware Horizon View to our user experience and application performance monitoring product. Read on for more goodness!

GPU acceleration is en vogue. After slowly but steadily moving out of the 3D niche it has arrived in the mainstream. Today, applications like Microsoft Office leverage the GPU, but even more so do web browsers. Chrome, Firefox, and Internet Explorer all have hardware acceleration turned on by default. People generally seem to be happy about that - GPUs are super-efficient, the more work they do the fewer remains for the CPU, overall energy consumption is reduced and battery life increases. Or so the myth goes. Interestingly, facts to prove that are hard to find. Nobody seems to have measured how GPU acceleration affects CPU usage. Let’s change that.

The most important change in version 2.0 of uberAgent is the new architecture. In addition to that we have made many other improvements, significantly reducing the agent’s footprint while increasing the quality of the collected data.

Note: The installation described in this article worked very well for me. However, nine years after publishing this post I replaced this setup with a more modern dockerized installation that is easier to maintain. PHP upgrades, especially, are a lot simpler.

Do you know the maximum number of users each of your terminal servers can host with acceptable performance? You may have found out the hard way how many are too many - but how many are just right? Farm sizing and server capacity planning are typical tasks for consultants who often have a hard time fighting the peculiarities of perfmon and logman trying to get the data they need for their calculations. It can be so much easier at no additional cost. The 60-day Enterprise Trial version of Splunk in conjunction with an evaluation license of uberAgent give all the information you need in much less time. Here is how.

When you move users from a physical PC to a VDI environment you may find that they are not too happy with their new machine’s performance - it happened to me. To quantify things I took a series of measurements comparing the old PCs we migrated away from with both VDI machines and the new PCs available to some.

In the last five years Xperf has gained popularity as an administrator’s secret weapon for battling all kinds of performance issues. But just when it was on the brink of becoming as mainstream as such a tool can be, Microsoft superseded it by something else: Windows Performance Recorder.

Today it is still as common as it was in the 90s to go for a coffee after switching on the PC in the morning, only to find it still booting when coming back several minutes later. SSDs have not arrived in the average office yet, and even they can only cure some of the diseases that cause computers to start up agonizingly slowly.

The days are long gone when a browser was just another application. Modern websites are applications of their own, and the browser is their operating system. That has consequences for monitoring. It is no longer sufficient to gather performance data for the browser as a whole. When, for example, Internet Explorer’s CPU usage is high, Administrators need to understand what caused that. Is it the business-critical ERP site or are people just watching fun videos on YouTube?

Considering the importance of fast logons for a good user experience there is surprisingly little information on the subject. Windows does not even record the total logon time, let alone where it is spent. Administrators wishing to analyze their users’ logons are left in the dark. But that can be changed.

[Update on 29/06/2012: added tests 5 and 6 with RAID controller cache and Atlantis Ilio] When a 400 GB SSD comes nearly at the price of a small car, you expect a lot, performance-wise. At least I did when I got my hands on two drives from HP’s enterprise performance series. Read on to find out if that hefty price tag is justified.

[What I wrote here is wrong. Please read this follow-up article.] Here is a very simple way to reduce the time it takes to boot Windows 7 by around 20% (the exact number depends on the hardware).

My website is hosted by Rochen, a company whose services I am very happy with. For reasons I understand they do not offer unlimited data transfer. My account, for example, comes with 50 GB per month. That is a lot and enough for most sites, including mine at this point in time. But I am selling software, and if a big news site writes about SetACL Studio the least of all things I want to happen is that Rochen suspends my account because the transfer limit has been reached.

Problem Firefox (3.6.13) intermittently freezes. This happens during scrolling, text input, basically everywhere in the user interface, rendering the browser nearly useless. Analysis While browsing amazon.de I created a log of Firefox’s activities by recording system activity for 85 seconds with Sysinternals Process Monitor. I then filtered the log to include only Firefox activities. Clicking on Tools -> Process Activity Summary I got:

This article is part of Helge’s Profile Toolkit, a set of posts explaining the knowledge and tools required to tame Windows user profiles. Redirecting AppData from the user profile to a folder on the network may significantly improve logon speed. Is enabling AppData redirection a no-brainer, then? Not really, because it often comes at a price: performance.

Normally, finding the cause for high CPU utilization is easy - just start Task Manager. But what if the component consuming CPU cycles is a driver that runs in the kernel? In that case, there is no regular process Task Manager could attribute the usage to. You will see the “System Idle Process” at around 98%, but the “Performance” tab might indicate 50% CPU usage. How can that be?

This article describes a bug that dramatically reduces file copy performance on Windows systems. I cannot provide a fix (not having access to the Windows source code), but I have found a workaround.

DiskLED is a small tool that displays performance counter data using an animated system tray icon. The new version 1.1 comes with a graphical configuration dialog that lets you choose from all performance counters available on the local system - you can monitor network throughput just as easily as hard disk activity or memory usage. In this article, I am going to show you how to do that.

What do you do when your computer reacts sluggishly to even the simplest commands? You probably look at its hard disk LED to determine if the disk is busy, because if it is, the only thing that really helps is waiting (apart from buying a faster disk or SSD). Problem solved - if you are sitting right next to the machine. But what if you are using a protocol like RDP or ICA to connect to a remote computer or VM? No HDD LED, no quick and simple way to check for hard drive activity. This has been bugging me enough to write a software replacement: DiskLED.

Here are some very simple guidelines that help determine how much memory a particular server actually needs in order to perform well. Of course, these are just rules of thumb, but nevertheless they might come in handy.