Help choosing replacement Hyper-V host machines and connected storage
-
@KOOLER said:
@scottalanmiller said:
@JohnFromSTL said:
Due to our size and budget constraints we purchase refurbished equipment, which are typically 2-3 years old. Whichever solution I end up with needs to last 3-4 additional years.
Refurb is good. @xByteSean is around to help out with that. As is @ryan-from-xbyte
xBytes allows people to get Dell quality @ SuperMicro price
That's a great way to look at it.
-
@scottalanmiller said:
@JohnFromSTL said:
@Minion-Queen said:
NTG can help with the Scale questions so ask away!
Any idea on starting price?
Did I miss your total memory needs? How much memory are you using added all up?
Host01:
CPU: (4) - Xeon X7350 @ 2.93GHz
RAM: (24) - 4GB DIMM 96GB RAM installed
RAID: PERC 5/i (RAID-5)
HDD: (5) - 300GB Hitachi Ultrastar 15K600 HUS156030VLS60 15k RPM 16MB Cache SAS 6Gb/s 3.5"Host02:
CPU: (4) - Xeon X7350 @ 2.93GHz
RAM: (24) - 4GB DIMM 96GB RAM installed
RAID: PERC 5/i (Msft Storage Spaces)
HDD: (5) - 1TB Seagate Constellation ES.3 ST1000NM0023 7200 RPM 128MB Cache SAS 6Gb/s 3.5" -
@KOOLER said:
Scale needs 3 nodes, they don't have anything to install on existing servers like OP has and Scale has no Hyper-V support.
Scale would be a change of direction. No need for John to manage his hypervisor or storage. Everything in a single package, ready to go.
-
It's good to see this thread with more legs over here on Mango than on SW.
Here is the interesting thing going on with the Dell generations now. We see the 11th gen still popular (R510s and R710s) if you only need 6-core CPUs. If you need 8-cores, then we see people skipping the 12th generation and going on to the 13th generation servers like the R730xd. The newer CPUs are so powerful at the low end and therefore so inexpensive, it is hard to justify other the 12th generation. There are some exceptions if you are okay with E5-2600 version 1 CPUs. That market is crashing and you can find some deals.
This is why we encourage people to call us so we can walk through the options in the secondary market. It is very different than buying directly through Dell.
-
@JohnFromSTL said:
RAID: PERC 5/i (RAID-5)
Well fixing that will be the first thing
We can do RAID 6 on the new servers. -
@ryan-from-xbyte said:
It's good to see this thread with more legs over here on Mango than on SW.
FTW!
-
@ryan-from-xbyte said:
The newer CPUs are so powerful at the low end and therefore so inexpensive, it is hard to justify other the 12th generation.
Except for when you are building storage devices. The R720xd really shines where CPU makes no difference.
-
@scottalanmiller said:
If the R910 is maxing out at, say, 20% CPU, then my guess is that an R720xd will do the trick to take over its load. The R720xd has two, faster procs than the R910. Not only are the individual procs faster, but by moving from quad procs to dual procs you gain a small amount of efficiency just from that one move. So faster procs and more efficient proc usage and then cutting the total number of procs in half.... seems like you will be okay.
You don't even need to go with a R720xd when comparing the R910. You can go with a R620 and get 10x2.5" drives. Cluster a couple of those together and you are far better off than getting an R910 and you will save a fortune on power.
-
@scottalanmiller said:
@JohnFromSTL said:
RAID: PERC 5/i (RAID-5)
Well fixing that will be the first thing
We can do RAID 6 on the new servers.Yeah, it was great when it was new...7 Years ago! I keep waiting for something to fail on these servers.
-
@JohnFromSTL said:
RAM: (24) - 4GB DIMM 96GB RAM installed
Matching or beating that will be easy. Even the older R720 / R720xd commonly came with 128GB. So no problem getting more than enough for you.
-
@StrongBad said:
@KOOLER said:
Scale needs 3 nodes, they don't have anything to install on existing servers like OP has and Scale has no Hyper-V support.
Scale would be a change of direction. No need for John to manage his hypervisor or storage. Everything in a single package, ready to go.
I cannot say better!
-
@scottalanmiller said:
@ryan-from-xbyte said:
The newer CPUs are so powerful at the low end and therefore so inexpensive, it is hard to justify other the 12th generation.
Except for when you are building storage devices. The R720xd really shines where CPU makes no difference.
If you don't need CPU performance and only need storage, I would recommend the R510. If you need the spindles, then the 24x2.5" R720xd would be a good fit. For raw capacity, the R510 will give you a cheaper option.
-
@ryan-from-xbyte said:
You don't even need to go with a R720xd when comparing the R910. You can go with a R620 and get 10x2.5" drives. Cluster a couple of those together and you are far better off than getting an R910 and you will save a fortune on power.
He's going to need LFF drives, I think, for this. He needs 16TB per node, from what I can tell, which is pretty big. If we have 2TB NL-SAS drives in RAID 6 that will require ten LFF bays just to hit the 16TB usable number.
-
@ryan-from-xbyte said:
If you don't need CPU performance and only need storage, I would recommend the R510. If you need the spindles, then the 24x2.5" R720xd would be a good fit. For raw capacity, the R510 will give you a cheaper option.
For storage capacity the R510 could definitely do it. I think that the R720xd with the 12x LFF drive option is probably where he needs to be. Enough drive capacity to do the 16TB usable and the improved CPU performance without breaking the bank.
-
@scottalanmiller said:
@ryan-from-xbyte said:
You don't even need to go with a R720xd when comparing the R910. You can go with a R620 and get 10x2.5" drives. Cluster a couple of those together and you are far better off than getting an R910 and you will save a fortune on power.
He's going to need LFF drives, I think, for this. He needs 16TB per node, from what I can tell, which is pretty big. If we have 2TB NL-SAS drives in RAID 6 that will require ten LFF bays just to hit the 16TB usable number.
I could probably shave 2-3TB off the SQL and Oracle VMs. I've included more just in case we need the space as databases are migrated from one version to another.
-
@JohnFromSTL said:
I could probably shave 2-3TB off the SQL and Oracle VMs. I've included more just in case we need the space as databases are migrated from one version to another.
You don't want to cut these things too closely.
-
One thing we have not considered is database storage performance. Yes we can get what we need by going with NL-SAS or even SATA (ugh) on RAID 6 and that will be pretty decent for read performance, but write performance will be really weak, especially for a database.
-
What about 10K SAS drives? Are any available that are big enough for that? What about 1.8TB 10K SAS?
-
Moving from 2TB drives up to 4TB RE drives won't hit 10K RPM per drive but the move from RAID 6 to RAID 10 will help a lot with database performance.
http://www.amazon.com/SAS-Enterprise-Hard-Drive-WD4001FYYG/dp/B0090UGQ2C
$203 for a 4TB RE SAS drive.
-
And for databases you will definitely want SAS over SATA, the access patterns heavily favour the SAS protocol. You could see as much as a 20% difference in performance between the protocols alone.
RAID 10 will double the write performance over RAID 6 as well for database writes, in nearly all cases. As these are probably going to be twelve bay, not ten bay, servers going to all twelve bays will improve storage performance too.
