Planning Server: HP, Dell, IBM, Cisco(?)
-
Having just read this thread and the article I have to say I didn't know that Cisco what making servers (maybe I did,.. just forgot).
But having had a few conversations with Scott and others, I wonder what my planning should be on server hardware... And what is HP Moonshot? The report didn't make it out to be a "take off" product line,.. but it is interesting.
As I am replacing 3 Dell PE servers, I wonder if I should stay with the PE line and Dell (via xByte) or bite the bullet and be more forward thinking..
Does MoonShot hold more 'options' over the PE line or is it just the scalability of it.....
-
Cisco makes servers but they are very complex and weird and generally require extra training. I've been forced to use them a few places and the general consensus has always been that while well made, the design is horrible and they should never be purchased. Any that I've seen deployed was a purchasing department being duped and the IT department being upset that something so complex, fragile and unnecessary was purchased without consulting them. I've yet to work in an environment that would buy them again.
-
Moonshot is for huge scale. It's when you need to deploy dozens or scores of servers (hardware) rapidly and need it in dense space and want flexible power consumption based on fluctuating load. Really not for the SMB market. It's for datacenters with high density. What makes Moonshot really awesome is if you meet all of those requirements and are able to recompile for the ARM platform and then it goes to another level of cool. Otherwise you are just running a large fleet of Intel Atom servers.
-
For an SMB the good options really come down to...
Dell PowerEdge R, HP Proliant DL, SuperMicro. Those are the only three that I would consider under most circumstances. All other products are either for large datacenter (Moonshot), for branch offices or other non-core business (VRTX), large scale grid (PowerEdge C) or similar.
-
@scottalanmiller said:
Moonshot is for huge scale. It's when you need to deploy dozens or scores of servers (hardware) rapidly and need it in dense space and want flexible power consumption based on fluctuating load. Really not for the SMB market. It's for datacenters with high density. What makes Moonshot really awesome is if you meet all of those requirements and are able to recompile for the ARM platform and then it goes to another level of cool. Otherwise you are just running a large fleet of Intel Atom servers.
Having done more than the curserory skim of the specs - I agree as I'm not going for a datacenter scale.. no need. cool yes... need no.
-
For a small business with small needs, the most common approach is just a single server. Unless you have medium or larger failover needs, a single, quality server is often the best. Few IT departments want to admit this because it makes IT seem less necessary than they want it to be, but in reality, a single server is extremely reliable and extremely cost effective. The cost of downtime is usually fractional compared to what people believe that it is, in SMBs it can often be nearly zero.
-
@scottalanmiller said:
For a small business with small needs, the most common approach is just a single server. Unless you have medium or larger failover needs, a single, quality server is often the best. Few IT departments want to admit this because it makes IT seem less necessary than they want it to be, but in reality, a single server is extremely reliable and extremely cost effective. The cost of downtime is usually fractional compared to what people believe that it is, in SMBs it can often be nearly zero.
This does contradict some of the main stream - however my SBS2003 server ran perfectly for years (don't know current state). It isn't an option currently for 1: obvious reasons- EOL; 2: I'm above the 75 user max limitation. But the SBS line was pretty reliable in the period of time I used it.
We are a mid side (maybe) SMB with about 300 users; of which about 75% are 'branch' users. We cover 14 counties with our main two business units. We have been moving more towards one single,.. but for the 'billing' they remain two separate... Only really matters in the OU to me.
-
@g.jacobse said:
...But the SBS line was pretty reliable in the period of time I used it.
And that was an old server, SBS and physical. All three contribute to unreliability. Going with a new server (new servers are more reliable than old one, the technology just keeps improving), avoiding SBS (it was just a collection of anti-practices mashed into a product) and adding virtualization takes an existing, working concept and improves it in three major ways. So if that was even remotely good previously, going with a single server of Windows Server 2012 R2 (or wait for Windows Server 2015) will be that much better. Both the hardware and the software have improved in many ways. It's a huge win for low cost.
Chances are that old server was not RAID 10 either. Switching to that might have a significant improvement in potential reliability as well. You can do a lot for reliability for relatively little money.
-
A second server, when not needed for performance or capacity reasons, is a major investment for an SMB. Often $5K to $15K it can be close to double the initial budget. And it adds a lot more than just up front hardware costs. There are licensing considerations and it adds some system complexity that does not exist with the single server approach creating more risk and cost (if in nothing else, in operational overhead.)
-
I've never used them personally* but xByte has a good rep in the community and what would appear to be excellent prices for used gear.
*Canada tax: the cost of additional shipping, border fees, import taxes, etc makes ordering hardware from the 'states just not worth the hassle. -
NTG uses xByte. They've been great. @g.jacobse already uses them so that works out well for him
