No announcement yet.

What Would You Like To See Next?

  • Filter
  • Time
  • Show
Clear All
new posts

  • #91
    Originally posted by Michael View Post
    There is an icon for each OS if you click on the test profile itself to view all the details... but yes, it might be wise adding it to that index page as well. When running PTS locally (e.g. phoronix-test-suite list-tests) by default it automatically shows only the supported tests relevant to that OS.
    Yeah. It is just that I would like to see at a glance what tests are supported on Windows and Linux. Such a feature would be useful for checking out which tests are cross-platform


    • #92
      Seriously Michael, do something to improve moderation on the forum. When there is a thread with a serious topic (OpenBSD runs out of money) trolled by the same person with several different accounts that seriously decreases the value of your forum.


      • #93
        Originally posted by Vim_User View Post
        Seriously Michael, do something to improve moderation on the forum. When there is a thread with a serious topic (OpenBSD runs out of money) trolled by the same person with several different accounts that seriously decreases the value of your forum.
        Volunteers wishing to be moderators are generally an honored request for those that have been long-time forum members, but sadly few step up.
        Michael Larabel


        • #94
          Benchmarks with real-world applicability.

          As a long-time reader, short time registered user, don't take this the wrong way, but it seems Phoronix focuses on graphics benchmarks, which are great and all, but the vast majority of other benchmarks (ie: file system, QEMU/KVM, etc... ) are virtually useless. Its just a bunch of numbers and graphs, with one line comments if we're lucky. It would be nice to see benchmarks that provide *real world* applicability and some sort of conclusion at the end based on the data. You do want people to read through the entire article across all its pages, well right now there is no reason too as it takes more clicks to get there than words in the conclusion. Some useful commentary at the end may help that.

          As for real-world applicability:

          - File systems:
          When it comes to file systems, showing the default settings of btrfs vs ext4 across one kernel version to another is essentially useless, especially when the performance differences are not noticeable to a normal user. Instead pick a common LTS kernel version, focus the articles on actual real-world use cases and how to improve them. Common use cases on how to get the most storage space for data, media, backups etc... This could focus on for example btrfs settings using deduplication (in-band or out-of-band), compression (different algos), and which settings may be worth while and which ones aren't. Don't just show numbers though, actually come to a conclusion based on the data and explain how the user can *replicate* the same settings on their own system. An article like this would last the test of time and be super useful to many people, in fact I would imagine many people would revisit it several times. Thats just one case though, then expand this out to other use cases like mail servers (high transaction rate, least disk usage), database servers (OLTP and data warehouse use cases for MySQL/PostgreSQL), the list is endless, but they would all be articles worth bookmarking and linking too by Q&A sites. Another nice thing would be bcache benchmarks vs DM-cache and their various settings.

          - Virtual Machines
          Cloud hosting is everywhere, and OpenStack itself is quite popular, articles detailing the fastest settings for CPU workloads, vs IO workloads (ie: different drivers, different image file formats or LVM partitions, different file systems (host and guest), etc...) as well as how you can improve IO for many guests on a single host using things like bcache. Again focus on real-world applicability that gives each article a focus, or purpose, like reducing IO latency with many guests, increasing density with smaller guests (overcommit vs undercommit), etc...

          The list goes on, but people are busy these days, they don't have time to spend days/weeks benchmarking things themselves with all the different variables that come into play, they want answers that are easy to find and follow. Thats why it would be great to a find an article that says: "Hey, if you want to setup a database server with the fastest IO, heres a great starting point, and heres a couple different settings to try.", or "Hey, you want to setup a home media server, heres a few settings to get the most out of your storage." Narrow down the variables to just a few, but have an actual primary focus/purpose for the article. Bonus points for detailed instructions on getting the test suite running for testing the exact same situations.

          Those would be articles worth reading, rather then kernel v3.89.19-rc183 is 0.0001% faster running BTRFS compared to v3.89.19-rc182, but 0.00003% slower with EXT4. I wish I could take back the time I've wasted clicking through an article where virtually every bar graph is within the margin of error of the next, quality over quantity please.