Cannot decide between 1U servers for growing company
-
@ntoxicator said:
Also with local storage.
Would still need PCIe 10GbE cards in each server host. And they inter-connect as "backplane" And this is how the data syncs, assuming DRBD
Ofcourse localized SAS 10k disks will be hella-lot faster than TCP/IP 10Gbe to a directly connected NAS/SAN
Yes, you will probably need 10Gbe.... or port bond a quad port gig adapter. Which may work just fine in your scenario. You would still need 10Gbe on your NAS if you wanted to get rid of that bottleneck.
-
Gotcha
Well the entire plan for the upgrades and future planning was to have 10Gbe all the way around. Have 10Gbe dual port cards on each server node.
and then dual port 10Gbe in the NAS device. Along with a 10Gbe Switch using SFP cables to interconnect. the servers & NAS would inter-connect to the 10Gbe Switch and be on their own IP network
-
@ntoxicator said:
Gotcha
Well the entire plan for the upgrades and future planning was to have 10Gbe all the way around. Have 10Gbe dual port cards on each server node.
and then dual port 10Gbe in the NAS device. Along with a 10Gbe Switch using SFP cables to interconnect. the servers & NAS would inter-connect to the 10Gbe Switch and be on their own IP network
So... why wouldn't you just do that with server nodes with local storage and get rid of a massive point of failure? The NAS device is introducing unnecessary risk to this setup and will be much slower and much less reliable then having the storage on board.
-
I'm still wondering why you need HA like this?
Sure every CEO thinks he can't afford downtime, but really 5 mins or even an hour is often well within the realm of acceptability when the dollars are added up.
-
One thing I'm not sure anyone has considered - do you really need more than one server? Can your entire load be handled from a single server?
-
Its a very fast pace office environment, ever one in the 'now' mentality. Any blip or downtime I have employee's bitching or CEO down my back to get it up.
So yes, wanted HA.
The NAS would be 2 of them running HA for centralized storage.
Otherwise, I can spec refurbished servers from xByte and be done with it
Just built a DELL R720
5 - 2TB 72000 RPM DELL certified hard drives (RAID-10, PERC H730)
2 - 50GB SLC Solid state drives - to install XenServer on - raid-1
750W Dell dual power supplies3 year onsite support/service
Right at 5,000 per server
-
@ntoxicator said:
Its a very fast pace office environment, ever one in the 'now' mentality. Any blip or downtime I have employee's bitching or CEO down my back to get it up.
So yes, wanted HA.
The NAS would be 2 of them running HA for centralized storage.
Otherwise, I can spec refurbished servers from xByte and be done with it
Just built a DELL R720
5 - 2TB 72000 RPM DELL certified hard drives (RAID-10, PERC H730)
2 - 50GB SLC Solid state drives - to install XenServer on - raid-1
750W Dell dual power supplies3 year onsite support/service
Right at 5,000 per server
Ok... for RAID-10 you will need an even number of drives that's just how RAID 10 works... why are you installing your hypervisor on SSDs? It is just loaded to RAM and those expensive disks are going to waste. Generally hypervisors should be run from the cheapest slowest disk you have since you are rarely writing or reading from them except on boot.
-
Wow - that performance is not going to be very good on the disk side of things.
Especially when you are looking to do VSAN.
-
I would rather have 2 servers in HA. I could most likely slam everything on one node.
each VM uses about 8GB of ram. (Domain controller uses under 4GB right now).
All our employee's connect to Terminal Servers.... due to 2X Application Gateway server..
-
@ntoxicator said:
5 - 2TB 72000 RPM DELL certified hard drives (RAID-10, PERC H730)
2 - 50GB SLC Solid state drives - to install XenServer on - raid-1
750W Dell dual power suppliesWhy are you splitting the drive types?
-
I'm going for storage array. You're the one who keeps pushing the localized storage.
As I would want to have over 4TB of localized storage to be on safe-side for future. As its locallized and CANNOT grow it
So at this point of localized storage. Appears HC Scale would be the kicker here.
-
Also making the assumption when running HA for localized storage (VSAN) or HA-Lizard iSCSI for XenServer.
The storage on each node will have to be the same so it can replicate on each node.
-
@ntoxicator said:
Its a very fast pace office environment, ever one in the 'now' mentality. Any blip or downtime I have employee's bitching or CEO down my back to get it up.
So yes, wanted HA.
This is a dollars and cents game. You design a properly sized single server that can handle your entire current load (maybe one that's over sized by say 10% so you have a little room for growth, unless you KNOW about growth that is coming, otherwise never plan for future 'talk').
Then also present the HA solution. Give estimates on downtime differences in cases of issues. Find out the real cost of down time to the company to discover what the RTO and RPO need to be.
My doctors all scream bloody murder at downtime, but when they consider the cost of a HA system, they are willing to live with the few outages we rarely have.
You're previous NAS solution, to put it bluntly, got lucky. You either never had an outage or you found quick solutions when you did. But what if you would have had a complete chassis failure on your NAS? You would have been completely down until it was fixed, and the recovery of that NAS might be more difficult than recovering traditional servers.
-
@Dashrender said:
@ntoxicator said:
Its a very fast pace office environment, ever one in the 'now' mentality. Any blip or downtime I have employee's bitching or CEO down my back to get it up.
So yes, wanted HA.
This is a dollars and cents game. You design a properly sized single server that can handle your entire current load (maybe one that's over sized by say 10% so you have a little room for growth, unless you KNOW about growth that is coming, otherwise never plan for future 'talk').
Then also present the HA solution. Give estimates on downtime differences in cases of issues. Find out the real cost of down time to the company to discover what the RTO and RPO need to be.
My doctors all scream bloody murder at downtime, but when they consider the cost of a HA system, they are willing to live with the few outages we rarely have.
You're previous NAS solution, to put it bluntly, got lucky. You either never had an outage or you found quick solutions when you did. But what if you would have had a complete chassis failure on your NAS? You would have been completely down until it was fixed, and the recovery of that NAS might be more difficult than recovering traditional servers.
You are correct.
We work on the back-office. We provide medical billing solutions to doctors offices (Hospital + and larger clients). Millions of billable dollars yearly billable to insurances.
CEO and team are anticipating adding another 30 employees by Q1 2016. Afterwards due to the clients in pipline, we will need additional 5-10 employees per month for remainer of 2016 year
Anticipated growth was 200 employees by end of 2016
-
And quite frankly its scary to think about if NAS chasis failure. yes all the VM's would be down and be entire SOS panic mode. eggs in one-basket at the moment. I've explained and shouted my concerns to CEO. Told him that it will be his responsibility if downtime as prior investment $$ was not there.
2013 year when installed it, I wanted dual servers + dual NAS. Pricing was only given for single units. There is an LSI raid card in the current Supermicro server and disk in raid-1 array. but thats a moot point
-
@ntoxicator said:
Also making the assumption when running HA for localized storage (VSAN) or HA-Lizard iSCSI for XenServer.
The storage on each node will have to be the same so it can replicate on each node.
When you're building a new HA system, you design it so everything is identical anyways to remove the need to "adjust" anything.
That is a moot point to mention.
Build an identical system for HA.
-
@Dashrender said:
I'm trying to understand the draw to more expensive Oracle/Cisco servers over HP or Dells?
Oracle are good, roughly equivalent.
Cisco is a big inferior. They are more like SuperMicro.
-
@Jason said:
@Dashrender said:
I'm trying to understand the draw to more expensive Oracle/Cisco servers over HP or Dells?
Oracle makes sense for Enterprise stuff.. Not as much as it used to. But you can't really compare them to your low end Dell/HP stuff.
Oracle Sparc gear is extremely high end and probably the best on the market. But their AMD64 gear is just on par with Dell and HP. Nothing special that I've seen.
-
@ntoxicator said:
I'm sure I could spec the servers enough with local storage to support the needs. However, then I be diving into HA-lizard terratory and would completely get away from our current NAS centralized storage.
If you need anything at SA (Standard Availability) or higher (like HA) you can't have that NAS in the picture. There is literally no means of using it (except as decoupled like as a backup target.) If you are comfortable with LA (Low Availability - meaning lower than a single server on its own with no protection of any sort) then you can use the NAS. But your stated goals rule the NAS out completely, no ifs ands or buts.
-
You can always get two SANs and Two Hosts but it will be really expensive for not much of a setup.
You can get server with well over 20TB of local storage.
