We Don't Have the Budget to Save Money
-
@scottalanmiller said:
The situation that just came up was someone who bought six "desktop class" servers without RAID cards for a non-profit and claimed that they could not afford hardware RAID because they had no money. Yes they could afford six servers when a single server, with a real RAID card, would have been safer, faster, cheaper and easier than the six servers that they bought. The problem of not having enough money was created because they wasted more money than was needed and then ran out.
Yeah, that situation sounds way too familiar. A similar excuse was made at a non-profit I used to work for when I brought up the need to begin virtualizing servers. I would always have to listen to the network manager complain about how hard their job was, how much they had to manually patch, backup, etc etc. The moment I talked about ESXi it was shot down so quickly because "Can't afford a SAN, you HAVE TO HAVE A SAN to do virtualization." and then did everything they could to torpedo the idea to upper management. Frustrating.
-
@NetworkNerd said:
I think the costs associated with more labor, more upkeep, more overall administration of a solution moving forward often get overlooked when decision makers begin to see how much money comes out of the bank account. The long-term costs are often not communicated to the big boys well, not communicated at all, or just flat out ignored.
That part is ALWAYS overlooked in my experience too. No one seems to want to factor those costs in or they conveniently leave those out so they can justify their lack of willingness to evolve in their positions.
-
@Bill-Kindle said:
"you HAVE TO HAVE A SAN to do virtualization."
I've heard this loads of times. It must be one of the biggest myths in IT. I'm not sure where it came from, the SAN vendors or somewhere else? Whatever, a lot of IT pros think it's true.
-
@Carnival-Boy said:
@Bill-Kindle said:
"you HAVE TO HAVE A SAN to do virtualization."
I've heard this loads of times. It must be one of the biggest myths in IT. I'm not sure where it came from, the SAN vendors or somewhere else? Whatever, a lot of IT pros think it's true.
All I know is that when I laughed at them for continuing to claim that, they became irate.
-
@Carnival-Boy said:
@Bill-Kindle said:
"you HAVE TO HAVE A SAN to do virtualization."
I've heard this loads of times. It must be one of the biggest myths in IT. I'm not sure where it came from, the SAN vendors or somewhere else? Whatever, a lot of IT pros think it's true.
This is the number one myth that was shattered for me when I joined SW. Thanks to Scott.
-
@Carnival-Boy said:
How much did they spend? I've bought HP's super-budget Proliant servers for less than $200 in the past. So six of those would come in at $1200. If you know nothing about virtualisation and would have to learn, then six budget servers could, in theory, work out cheaper. And how unreliable is HP's software RAID, anyway?
In this case it was not HP, it was Dell. Same difference, just saying. But Dell does not have the equivalent of the $200 HP MicroServer. Their entry level is basically a $600 - $800 desktop unit. There is no way that six of them was anywhere nearly as cheap as a single slightly better unit with hardware RAID.
And the software RAID not working is what brought the topic up. Third party awkward software RAID that both HP and Dell offer is not something you would ever want to run in production, maybe a lab but even then I would shy away from it. It's not Fake RAID, but it isn't integrated OS RAID either. It is still a third party driver with compatibility risks and relatively little support as no serious business uses it - it's the "red headed stepchild" of RAID solutions so while it is generally supported, it is not supported well and gets little attention. It's really a "failing SMB" solution only, not something that the vendor's profit making customers are using so little reason for them to really care.
But even if in theory the software RAID solution was good in this case (there is a reason why this only exists for Windows and if you were running any other OS on the same hardware that you would use the OS RAID instead) it was specifically the inability to get the software RAID to keep working and to get support for it that created the discussion.
-
Again, in theory, I could imagine a situation where an IT guy needs new servers and has a tiny budget. He thinks about virtualistation, but doesn't really know much about it. So he contacts a couple of IT vendors and asks them to quote. All their quotes contain SANs and are thus eye-wateringly expensive. Some of them tell him he needs TWO SANs for redundancy. If the cost of buying two SANs doesn't put him off, the diagram of the swiches would - there's cables everywhere. And the labour just to set the SANs up! I can see why he might think virtualisation just isn't an option.
Now I know you'll say that's why you shouldn't get advice from a vendor, but paid, independent advice isn't cheap either. This is where Spiceworks rocks, I get tons of advice from an NTG consultant and I don't pay a penny for it, but not everyone knows about that! Generally, getting advice from forums can be extremely hit and miss, and you need to be fairly experienced to tell the expert advice from the bullshit - not everyone is.
-
@Carnival-Boy said:
Now I know you'll say that's why you shouldn't get advice from a vendor, but paid, independent advice isn't cheap either.
$200 is cheap compared to wasting thousands in up front purchasing, being at risk from not assessing needs and putting the company at product risk for the entire server cycle until it is refreshed, though. Good paid consulting for a tiny firm isn't necessary that expensive. You just don't command the schedule as much at that price (not like you are scheduling a full day.)
-
@Carnival-Boy said:
This is where Spiceworks rocks, I get tons of advice from an NTG consultant and I don't pay a penny for it, but not everyone knows about that! Generally, getting advice from forums can be extremely hit and miss, and you need to be fairly experienced to tell the expert advice from the bullshit - not everyone is.
And MangoLassi (cough, cough). But yeah, there are tons of awesome consulting firms and consultants who give advice freely in the public space. There are many options to getting advice out there when it is needed.
-
This mentality is way too common. Anything that people do not know is labeled as "too expensive." And then shops who don't hire good IT talent wonder why everything that they do costs so much. There is a reason that good IT people cost real money, because they save real money (or drive innovation.) Not every company needs expensive IT in house all of the time, but everyone needs it now and then.
-
@scottalanmiller Speaking of which, here's another example:
-
-
@Bill-Kindle said:
"Can't afford a SAN, you HAVE TO HAVE A SAN to do virtualization." and then did everything they could to torpedo the idea to upper management. Frustrating.
Wow. I've never actually used my SANs for Virtualization.. Just don't like the eggs all in one basket as they say.
-
@ajstringham said:
@Bill-Kindle said:
@scottalanmiller Speaking of which, here's another example:
Wow...
It's a rough one. I had not seen it until it was linked here.
-
You see this behavior a lot more in non-profits. I assume that this has to do with the lack of management training that is had there. People with MBAs or extensive experience are expensive. When a non-profit cuts corners that often means cutting corners in the places that would save them money, like management that understands money. So the result is losing money all over the place and no one realizing it or caring.
-
@scottalanmiller said:
@Carnival-Boy said:
How much did they spend? I've bought HP's super-budget Proliant servers for less than $200 in the past. So six of those would come in at $1200. If you know nothing about virtualisation and would have to learn, then six budget servers could, in theory, work out cheaper. And how unreliable is HP's software RAID, anyway?
In this case it was not HP, it was Dell. Same difference, just saying. But Dell does not have the equivalent of the $200 HP MicroServer. Their entry level is basically a $600 - $800 desktop unit. There is no way that six of them was anywhere nearly as cheap as a single slightly better unit with hardware RAID.
And the software RAID not working is what brought the topic up. Third party awkward software RAID that both HP and Dell offer is not
I tossed it about a time or two,.. Software RAID or hardware - Even with my lack of knowledge and hand on experience with RAIDs, I'd go hardware.. just seems safer and more reliable.
-
@g.jacobse said:
I tossed it about a time or two,.. Software RAID or hardware - Even with my lack of knowledge and hand on experience with RAIDs, I'd go hardware.. just seems safer and more reliable.
I think what a lot of people miss is that it is more reliable when humans are involved. Good software RAID (like that built into Linux or Solaris) is rock solid. But it requires more knowledge and planning than hardware RAID does, at least the good stuff. A SmartArray or a PERC is going to "just work." You don't even need a system admin. Your server tech in the server room can swap drives when he sees an amber light come on, no need to grab the system admin for such a low level hardware task.
Put software RAID in there and that same drive swap could be a disaster and the system admin has to coordinate the swap. That blind swap capability and automated rebuilds allow for a separation of duties and an abstraction from the OS that is really critical in the SMB and is handy anywhere.
Software RAID can be great. But hardware RAID has some real advantages.
-
@scottalanmiller said:
@g.jacobse said:
I tossed it about a time or two,.. Software RAID or hardware - Even with my lack of knowledge and hand on experience with RAIDs, I'd go hardware.. just seems safer and more reliable.
I think what a lot of people miss is that it is more reliable when humans are involved. Good software RAID (like that built into Linux or Solaris) is rock solid. But it requires more knowledge and planning than hardware RAID does, at least the good stuff. A SmartArray or a PERC is going to "just work." You don't even need a system admin. Your server tech in the server room can swap drives when he sees an amber light come on, no need to grab the system admin for such a low level hardware task.
Put software RAID in there and that same drive swap could be a disaster and the system admin has to coordinate the swap. That blind swap capability and automated rebuilds allow for a separation of duties and an abstraction from the OS that is really critical in the SMB and is handy anywhere.
Software RAID can be great. But hardware RAID has some real advantages.
Wouldn't hardware RAID be faster?
-
@ajstringham said:
Wouldn't hardware RAID be faster?
Not since around 2001 or so. It is actually slower. There is more horsepower in the main CPUs than there is in the embedded processors even on the best RAID cards. And the best cards only have a 1GB cache whereas software RAID has as much cache as you want to throw at it, 2GB is easy, 32GB isn't unheard of.
The idea that hardware RAID is faster comes from the Pentium 32bit era (1990s) when hardware RAID was new. Back then the main CPU(s) were so slow and overloaded that offloading the parity calculations was a big deal. It was in the Pentium Pro era (inclusive of the PPro, P2, P3 and Core processors) that the CPU(s) got so fast and were so rarely overloaded that software RAID passed hardware RAID in performance. So that could have been in the 1990s but was more reasonably in the early 2000s. Big systems like Power and Sparc systems never went to hardware RAID at all (and were always the fastest systems) because they always had stable operating systems with super fast CPUs and more of them.
-
@scottalanmiller said:
@ajstringham said:
Wouldn't hardware RAID be faster?
Not since around 2001 or so. It is actually slower. There is more horsepower in the main CPUs than there is in the embedded processors even on the best RAID cards. And the best cards only have a 1GB cache whereas software RAID has as much cache as you want to throw at it, 2GB is easy, 32GB isn't unheard of.
The idea that hardware RAID is faster comes from the Pentium 32bit era (1990s) when hardware RAID was new. Back then the main CPU(s) were so slow and overloaded that offloading the parity calculations was a big deal. It was in the Pentium Pro era (inclusive of the PPro, P2, P3 and Core processors) that the CPU(s) got so fast and were so rarely overloaded that software RAID passed hardware RAID in performance. So that could have been in the 1990s but was more reasonably in the early 2000s. Big systems like Power and Sparc systems never went to hardware RAID at all (and were always the fastest systems) because they always had stable operating systems with super fast CPUs and more of them.
Interesting. Good to know.
