The Software RAID Inflection Point
-
aww, so that it. The price of a RAID controller is so low that getting a software solution isn't worth the effort. OK I think I'm on the same page now.
They can't do it cheaper, and still make money.. so why bother doing it at all. That makes sense.
-
@Dashrender said in The Software RAID Inflection Point:
aww, so that it. The price of a RAID controller is so low that getting a software solution isn't worth the effort. OK I think I'm on the same page now.
They can't do it cheaper, and still make money.. so why bother doing it at all. That makes sense.
Right, they don't care because it isn't much money and hardware RAID is universal (doesn't care which OS or hypervisor is used on top). If they need to pay extra for software RAID, the benefits diminish. And if you really need to save money, you can use MD RAID already today. Or you can get hardware RAID refurbished and it is not all that much, maybe $400. So the total amount of savings for customers is tiny, the effort is bigger, there is more risk and the potential profits to a vendor are super tiny, especially as RAID is basically dead already, and no direct way for any vendor to capitalize on it.
So no one is looking to pay for this solution, and no one is looking to make money by creating it. There is neither an itch to be scratched nor a scratch looking for an itch.
-
What does @scale use? RAIN?
-
@Dashrender said in The Software RAID Inflection Point:
What does @scale use? RAIN?
Yes. Must like I describe with an additional layer that understands the drive speeds (like SSD, 10K and 7.2K.) So it does the things I mention plus it knows which bits go onto which type of storage.
-
You can't have any kind of effective hyperconvergence without RAIN. RAID doesn't scale out well at all. You can use RAID under RAIN, but that is pretty limiting.
-
@scottalanmiller said in The Software RAID Inflection Point:
Here is the thing, I think you are scratching an itch that just isn't there. Who actually needs this functionality? Who is the real world use case where this script would change buying decisions? And how do those changed buying decisions turn into direct financial benefit for the company that makes and supports the new software?
I'll just throw a quick thank you to @Dashrender, as almost everyone in IT seems to go through this same thought process at some point. Now we have a reference page
-
@Dashrender said in The Software RAID Inflection Point:
@scottalanmiller said in The Software RAID Inflection Point:
@Dashrender said in The Software RAID Inflection Point:
@scottalanmiller said in The Software RAID Inflection Point:
@Dashrender said in The Software RAID Inflection Point:
The hardware RAID is already taking away the work from MSPs / ITSPs, so what's the difference there?
It doesn't work that way. It takes no time to set up either. MD RAID is already a "zero effort" set up process. But there is pay for supporting it. It's not that this would really hurt an MSP, but it would cost a fortune to make it and support the code, money that could never be recouped, not even with tens of thousands of customers all using it.
So each Linux distro has it's own version of MD? it's not universal like so many other components?
How did you get from somewhere to this?
From here.
@scottalanmiller said in The Software RAID Inflection Point:
@Dashrender said in The Software RAID Inflection Point:
Who pays for MD development now?
The OS vendors, of course. It's part of the OS.
It's actually part of the Linux Kernel now. So maintained under that for all Linux distributions.
-
@travisdh1 said in The Software RAID Inflection Point:
It's actually part of the Linux Kernel now. So maintained under that for all Linux distributions.
Has been for a long time. But the majority of the Linux kernel is maintained by Red Hat, IBM, Suse and Canonical.
-
So it's not maintained by the OS vendors - it's maintained by the kernel support team?
-
@Dashrender said in The Software RAID Inflection Point:
So it's not maintained by the OS vendors - it's maintained by the kernel support team?
Which is mostly from the OS vendors. Who do you think makes the kernel?
-
@scottalanmiller said in The Software RAID Inflection Point:
@Dashrender said in The Software RAID Inflection Point:
So it's not maintained by the OS vendors - it's maintained by the kernel support team?
Which is mostly from the OS vendors. Who do you think makes the kernel?
Whoever Linus Torvalds thinks submits a good kernel patch. Which today is mostly the OS vendors.
-
@travisdh1 said in The Software RAID Inflection Point:
@scottalanmiller said in The Software RAID Inflection Point:
@Dashrender said in The Software RAID Inflection Point:
So it's not maintained by the OS vendors - it's maintained by the kernel support team?
Which is mostly from the OS vendors. Who do you think makes the kernel?
Whoever Linus Torvalds thinks submits a good kernel patch. Which today is mostly the OS vendors.
https://www.redhat.com/en/about/blog/red-hat-leads-open-source-contributions-to-kernel
-
Red Hat, Suse and IBM do the majority of the work with Intel as an "also mentioned."
-
By 2011 this had changed some, Novell/Suse had fallen to fourth and Intel picked up the pace. Microsoft was at 17th at that point.
-
Canonical is probably somewhere up there in the top ten. But over time it remains that Red Hat, Suse, IBM and Intel are the big four that make Linux happen. And overall corporate is 75% of the kernel while volunteer is 25%.
Big banks, governments and military are all big contributors as well.
-
While your statement doesn't imply that the MD versions are different, it does lead one who's not knowledgeable to a false conclusion that each Linux distro (at least the base distros) are maintaining their own version of MD. According to Travis, since it's backed into the kernel, that's hopefully not the case.
Basically what I was trying to determine, is MD pretty much the same across all distros of Linux?
-
@Dashrender said in The Software RAID Inflection Point:
While your statement doesn't imply that the MD versions are different, it does lead one who's not knowledgeable to a false conclusion that each Linux distro (at least the base distros) are maintaining their own version of MD.
Why would it lead someone to that? There is only one Linux kernel, where would different versions come from and where was that implied?
-
@Dashrender said in The Software RAID Inflection Point:
Basically what I was trying to determine, is MD pretty much the same across all distros of Linux?
Linux is Linux, MD is just part of it. There is one Linux kernel, one MD. Where is the idea that there are different Linux or MD or other components coming from? Honest question, what makes the assumption that a singular thing would naturally vary, happen? I saw nothing that would have led someone to think that, so I'm lost.
-
By 2015, corporate is now 80% of the Linux kernel development. Here is the latest contributions list: The Top 10 organizations sponsoring Linux kernel development since the last report include Intel, Red Hat, Linaro, Samsung, IBM, SUSE, Texas Instruments, Vision Engraving Systems, Google and Renesas.
-
@scottalanmiller said in The Software RAID Inflection Point:
@Dashrender said in The Software RAID Inflection Point:
While your statement doesn't imply that the MD versions are different, it does lead one who's not knowledgeable to a false conclusion that each Linux distro (at least the base distros) are maintaining their own version of MD.
Why would it lead someone to that? There is only one Linux kernel, where would different versions come from and where was that implied?
You said it's supported by the OS vendors - since so many other OS vendors don't work together, why would I assume they would be working together.
A better response originally from you which would have stopped this whole thread of conversation would have been to say - it's supported by many OS vendors through their cooperative work on the kernel, where MD is supported natively now. or some such statement. Instead we're 30 posts later and only now getting that specific knowledge from probing questions, where the probes are made out to appear like ludicrous considerations.
