Исходное сообщение
"Исследование поведения Linux на SSD-накопителе" Отправлено Michael Shigorin guest, 22-Сен-11 10:18
> Вот автора статьи я тоже уважаю. Статья действительно слабенькая, как и комментарии... накатал с разбегу: CFQ has had its overly fair share of troubles back then, mind you — I’ve stuck with deadline and was much happier. Re XFS, two things: 1) the current behaviour regarding “unsure” blocks at crash recovery is “deploy zerogun in case there’s a sensitive info leak” (was supposed to change but dunno if they’ve done it as of 3.0.4 — it’s rather a local policy decision, not fs design decision, and should be tweakable); 1a) you do want stable hardware, stable kernel+drivers and stable power (including physical connector safety) with XFS to avoid losing data; 1b) don’t call me names, I’m using XFS since ca. 2003 for storage including multi terabyte by now; 2) XFS had pretty bad CPU wakeup footprint (which does matter for portables) a year or two ago, while ext3/ext4 didn’t (even if they tend to have much worse RAM and disk space overhead). XFS is magnificent under heavy load, especially mixed/scattered one — provided one did take care of the aforementioned. Re SSDs, the article is pretty weak: 1) no mention of partition alignment which is *critical* for performance and durability of both SSDs and 4k-sectored HDDs (and erase block size issues not covered at all); 2) no mention of “discard” and “noatime” mount options. Part 2 is 404 by now, and I’m too lazy to dig up it on web.archive.org. :)

Ваше сообщение
Имя*:
EMail:	Для отправки ответов на email укажите знак ! перед адресом, например, !user@host.ru (!! - не показывать email). Более тонкая настройка отправки ответов производится в профиле зарегистрированного участника форума.
Заголовок*:
Сообщение*:	>> Вот автора статьи я тоже уважаю. > Статья действительно слабенькая, как и комментарии... накатал с разбегу: > CFQ has had its overly fair share of troubles back then, mind > you — I’ve stuck with deadline and was much happier. > Re XFS, two things: > 1) the current behaviour regarding “unsure” blocks at crash recovery is “deploy > zerogun in case there’s a sensitive info leak” (was supposed to > change but dunno if they’ve done it as of 3.0.4 — > it’s rather a local policy decision, not fs design decision, and > should be tweakable); > 1a) you do want stable hardware, stable kernel+drivers and stable power (including > physical connector safety) with XFS to avoid losing data; > 1b) don’t call me names, I’m using XFS since ca. 2003 for > storage including multi terabyte by now; > 2) XFS had pretty bad CPU wakeup footprint (which does matter for > portables) a year or two ago, while ext3/ext4 didn’t (even if > they tend to have much worse RAM and disk space overhead). > XFS is magnificent under heavy load, especially mixed/scattered one — provided one > did take care of the aforementioned. > Re SSDs, the article is pretty weak: > 1) no mention of partition alignment which is *critical* for performance and > durability of both SSDs and 4k-sectored HDDs (and erase block size > issues not covered at all); > 2) no mention of “discard” and “noatime” mount options. > Part 2 is 404 by now, and I’m too lazy to dig > up it on web.archive.org. :)
	Введите код, изображенный на картинке: