David R. Meyer wrote:
>>From the standpoint of the customers I work with on a daily basis, I'd
>say about 80% are using HW RAID for the reasons listed above. Those
>with SW RAID who had experience with HW RAID preferred HW RAID. Never
>had someone tell me that they preferred SW RAID.
>
>
I prefer software RAID. Strongly.
Cost is the primary motivation, followed closely by my like for lower
level software control of hardware.
Hardware RAID _does_ make sense for customers with:
1. Deep pockets.
2. Love of hardware vendor reliance.
3. Need for low CPU overhead.
When you're paying for Hardware RAID, you're usually paying for
redundantly built servers with high-cost SCSI/SAS drives (more platters,
lower bit densities, much better MTBF). When the hardware fails in some
obscure way, you call the vendor's tech in to fix it (and you're down in
the interim if you didn't architect your software to operate across
multiple servers in a redundant manner). If you have loads of cash
handy, you might have a cold/warm standby or some spare hardware to swap
out to get your expensive server back up and running.
Commodity computing looks at the problem differently: if you architect
the software solution to a critical computing system in a way so that
all nodes are stateless (or otherwise replicated) so that any hardware
outages are transparent to the application users, you no longer need
expensive hardware. Plus, you have far more resource capacity to scale
your application per $ spent.
You buy one big redundant server. I'll take 8 commodity PCs. I'll have
more space, more memory, and more raw CPU power, but I'll need to build
my software differently to take advantage of it.
That's my take on hardware RAID. For that matter, that's my take on most
things hardware centric.
A hardware solution to a problem is generally more expensive than a
software one. That is to say a hardware architected solution is usually
more expensive to a software architected one. It is usually cheaper to
throw commodify hardware with custom software at a problem rather than
build a unique piece of hardware to solve that particular problem.
For a fun read, check out this blog post:
http://blogs.sun.com/roller/page/jonathan?entry=the_rise_of_the_general
Custom ASICs are losing ground to more general software solutions to
problems, and that makes a software guy like me _very_ excited.
It's all about living with the economics...
- Ian C. Blenke <ian@blenke.com> http://ian.blenke.com
-----------------------------------------------------------------------
This list is provided as an unmoderated internet service by Networked
Knowledge Systems (NKS). Views and opinions expressed in messages
posted are those of the author and do not necessarily reflect the
official policy or position of NKS or any of its employees.
This archive was generated by hypermail 2.1.3 : Fri Aug 01 2014 - 20:41:15 EDT