Announcement

Collapse
No announcement yet.

Linux 6.9 Will Boot Much Faster For Systems With Large Amounts Of RAM

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • kylew77
    replied
    Originally posted by Paradigm Shifter View Post
    I've got a system with 1.5TB, and we are discussing upgrading it to either 3TB or 6TB for a project... and I'll be honest, booting Linux is not the reason for slow boot with 1.5TB. The issue is all the pre-POST stuff the motherboard does before it even outputs an image, closely followed by the RAID controller... and that isn't even our slowest server; our dual-socket Epyc takes forever to boot. Once the UEFI hands over to the OS, I've got SSH access in less than 15 seconds.
    We had Enterprise grade Dell EPYC and Xeon servers at my last work and they took forever to POST so I get what you are saying!

    Leave a comment:


  • Paradigm Shifter
    replied
    I've got a system with 1.5TB, and we are discussing upgrading it to either 3TB or 6TB for a project... and I'll be honest, booting Linux is not the reason for slow boot with 1.5TB. The issue is all the pre-POST stuff the motherboard does before it even outputs an image, closely followed by the RAID controller... and that isn't even our slowest server; our dual-socket Epyc takes forever to boot. Once the UEFI hands over to the OS, I've got SSH access in less than 15 seconds.

    Leave a comment:


  • uxmkt
    replied
    Originally posted by kylew77 View Post
    I frankly didn't even know 12TB of RAM was possible in a single server!
    SGI UV 300 supports 64 TB of RAM. One can debate whether that qualifies as a "single server", but it's a single server to the Linux kernel.

    Leave a comment:


  • Alliancemd
    replied
    My first reaction was "Ow, I have 'Large Amounts Of RAM`"(I have 64GB), read until the word "HugeTLB" and realized, wait, I don't have "Large Amount Of RAM"
    There are people out there with serious servers, not my pesky PC.

    Leave a comment:


  • byteabit
    replied
    Originally posted by niner View Post
    To be fair, nobody has said this, ever.
    Wasn't it "just" a limitation of the hardware/software from that time? I don't think that any smart person from that era was thinking it would be enough forever. Maybe a marketing person would say that or someone who don't understood computers. It's not impossible that someone "could" have made this statement.

    Leave a comment:


  • niner
    replied
    Originally posted by Jumbotron View Post
    I remember someone saying that I would never need more than 640K of RAM....<ahem>.
    To be fair, nobody has said this, ever. The quote you're alluding to is unlikely to be an actual quote. But even if it were, "640K ought to be enough for anybody" does not say "forever". It would have been an observation valid for that time. And at that time 64K was the norm, so 640 was indeed, quite luxurious.

    Leave a comment:


  • aviallon
    replied
    Originally posted by dragorth View Post

    You will get to see it at the replacement time.
    Unless you want to try an champion the use of an out of band Kernel to increase up time. I am sure there is a business case that could be made, though I have no idea on how much money would be saved on downtime for that half a minut.
    That's why I prefer using kernels from Oracle (Unbreakable Kernel) or upstream lts kernels directly.

    Leave a comment:


  • byteabit
    replied
    Originally posted by MorrisS. View Post
    So what is it meant for large amount of RAM? What's the impact on 16 or 32 GB or even 64?
    None. The improvements here are only relevant to really huge amounts of RAM, such as in TB range.

    Leave a comment:


  • MorrisS.
    replied
    So what is it meant for large amount of RAM? What's the impact on 16 or 32 GB or even 64?

    Leave a comment:


  • dragorth
    replied
    Originally posted by Rob72 View Post
    We have 21 servers with 4TB RAM, so this would be nice for them, because the bootup time is noticeably longer on them, compared to our standard servers with "only" 512GB. But realistically we will never see the benefit on these servers, as they are running Rocky Linux 8.x (kernel 4.18), and are unlikely to be upgraded even to Rocky Linux 9.x (kernel 5.14) before they are replaced, never mind a 6.9 kernel.

    But such memory amounts will become more common, so we should get to see the benefit at some point with future servers.
    You will get to see it at the replacement time.
    Unless you want to try an champion the use of an out of band Kernel to increase up time. I am sure there is a business case that could be made, though I have no idea on how much money would be saved on downtime for that half a minut.

    Leave a comment:

Working...
X