Proliant buying advice
-
So if "Enterprise Value" drives have massive IOPS, massive reliability and massive life expectancy, who buys "Enterprise Mainstream" and "Enterprise Performance" drives, given the massive premiums you're paying for the latter?
Is it possibly a bit of a money-making scam from HP to convince companies to overspend, or am I missing something else?
-
@Carnival-Boy said:
So if "Enterprise Value" drives have massive IOPS, massive reliability and massive life expectancy, who buys "Enterprise Mainstream" and "Enterprise Performance" drives, given the massive premiums you're paying for the latter?
We have no idea what those are.
-
@Carnival-Boy said:
Is it possibly a bit of a money-making scam from HP to convince companies to overspend, or am I missing something else?
Very possible.
-
@scottalanmiller said:
Each port is another full SAS channel with full SAS bandwidth. You want that if you have a place to use it. You don't, so does not matter to you. It is often used when you have an internal array and an external array or a massive internal array that you want to split.
From the specs, the 440ar has 2 internal 4 mini-SAS connectors, whilst the 440 has a single internal 8 mini-SAS connector. So both have 8 connectors, but the ar is split into 2 x 4 for whatever reason (form factor?).
-
@Carnival-Boy said:
@scottalanmiller said:
That's a separate issue. That's not the SSD wearing out, that's HP replacing drives proactively. As long as you are under warranty, that's a free, new drive.
I don't think so if this thread is anything to go by:
http://community.spiceworks.com/topic/465098-dell-enterprise-ssd-drives
It's Dell, but I'd guess HP is the same.Confirmed by HP: "Maximum usage limit: This is the maximum amount of data that can be written to the drive. Drives that have reached this limit will not be eligible for warranty coverage.". So whilst normal HDDs will last forever (under the server's warranty), SSDs won't. But this might be irrelevant if no-one ever reaches the maximum usage limit.
-
The cheapest SSD drives are SATA. Now with normal HDDs I think I understand the limitations of SATA - 7.2k spindle speed and less reliability (lower MTBF). But I can't find any distinctions between SAS and SATA SDDs other than throughput - 6G SATA versus 12G SAS.
Is throughput likely to be a constraint?
Are there any other differences?
-
@Carnival-Boy said:
@Carnival-Boy said:
@scottalanmiller said:
That's a separate issue. That's not the SSD wearing out, that's HP replacing drives proactively. As long as you are under warranty, that's a free, new drive.
I don't think so if this thread is anything to go by:
http://community.spiceworks.com/topic/465098-dell-enterprise-ssd-drives
It's Dell, but I'd guess HP is the same.Confirmed by HP: "Maximum usage limit: This is the maximum amount of data that can be written to the drive. Drives that have reached this limit will not be eligible for warranty coverage.". So whilst normal HDDs will last forever (under the server's warranty), SSDs won't. But this might be irrelevant if no-one ever reaches the maximum usage limit.
Ah, just a sales tactic then. Has nothing to do with the drives wearing out.
-
@Carnival-Boy said:
The cheapest SSD drives are SATA. Now with normal HDDs I think I understand the limitations of SATA - 7.2k spindle speed and less reliability (lower MTBF). But I can't find any distinctions between SAS and SATA SDDs other than throughput - 6G SATA versus 12G SAS.
Is throughput likely to be a constraint?
Are there any other differences?
Throughput is constrained but its not bad once you hit 6Gb/s. But the ability to use 12Gb/s is nice for a memory device.
-
@scottalanmiller said:
@Carnival-Boy said:
The cheapest SSD drives are SATA. Now with normal HDDs I think I understand the limitations of SATA - 7.2k spindle speed and less reliability (lower MTBF). But I can't find any distinctions between SAS and SATA SDDs other than throughput - 6G SATA versus 12G SAS.
Is throughput likely to be a constraint?
Are there any other differences?
Throughput is constrained but its not bad once you hit 6Gb/s. But the ability to use 12Gb/s is nice for a memory device.
Do you think his situation would even come close to saturating 6 Gb/s?
I suppose it could if he ran a very intensive query... but look at my question more generically. -
An SSD can pump a lot of data very quickly. Remember that we are talking about a memory bus here, not something talking to spindles. So pretty typically an SSD can saturate 6Gb/s. Even a single one.
-
Hmmn, I think I may stay old-skool and stick with 16 x 300gb SAS disks.
-
Update for those that are interested:
I've compromised in the end and gone with 12 x 300gb storage. This gives slightly less storage than I was hoping for, but means the IOPS is at least as good as we have now. I've also treated myself to a bit extra memory, since it's pretty cheap, so will have 96GB. The ERP system will be given 64GB of this (the max SQL Server standard allows), which will give loads of room to fit the entire database into memory, so there will be very little disk activity anyway.
I'll replace our other two Proliants next year, and may replace them with a single server. So we'll eventually go from 3 hosts down to 2. Although I'm not convinced that two hosts is better than three since disks are the most expensive part and you need more storage with a two host solution (for redundancy).
Quick question - since our ESXi licence only supports 3 hosts, and we have 3 hosts, how easy is it to replace one of the hosts? Will I able to add the new host prior to removing the old host (which would mean temporarily having 4 hosts)?
-
@Carnival-Boy said:
Update for those that are interested:
I've compromised in the end and gone with 12 x 300gb storage. This gives slightly less storage than I was hoping for, but means the IOPS is at least as good as we have now. I've also treated myself to a bit extra memory, since it's pretty cheap, so will have 96GB. The ERP system will be given 64GB of this (the max SQL Server standard allows), which will give loads of room to fit the entire database into memory, so there will be very little disk activity anyway.
I'll replace our other two Proliants next year, and may replace them with a single server. So we'll eventually go from 3 hosts down to 2. Although I'm not convinced that two hosts is better than three since disks are the most expensive part and you need more storage with a two host solution (for redundancy).
Quick question - since our ESXi licence only supports 3 hosts, and we have 3 hosts, how easy is it to replace one of the hosts? Will I able to add the new host prior to removing the old host (which would mean temporarily having 4 hosts)?
Nice work
What I would do is move the vm's off the host between the other two hosts (providing you have the capacity) , remove the license and then add the new host, license it and move the VM's to the new server.If you don't have the capacity to do that, it gets a bit more complicated.
-
@nadnerB said:
If you don't have the capacity to do that, it gets a bit more complicated.
Not really. You forget that you can install a new copy of ESXi and use it unlicensed with full features.
That should be plenty of time to migrate things.
-
Good idea, thanks.
-
@Carnival-Boy said:
I'll replace our other two Proliants next year, and may replace them with a single server. So we'll eventually go from 3 hosts down to 2. Although I'm not convinced that two hosts is better than three since disks are the most expensive part and you need more storage with a two host solution (for redundancy).
What do you mean for Redundancy? Why would I need more storage for two hosts vs three hosts?
-
Because I make sure I have enough storage so that if one host fails, I can move the VMs onto the other host(s). So with two hosts, I have to keep 50% of the storage free for failover. With three hosts, I only need to keep 33% of the storage free. So I'm getting up to 66% disk utilisation with 3 hosts, versus 50% with 2 hosts. Kind of vaguely similar to RAID5 vs RAID1.
To put it another way, with two hosts a host failure results in a 50% loss of resources, whilst with three hosts it results in a 33% loss of resources.
-
Makes sense.
