Moving to local storage?
-
Well that's fine an optical drive is seldom used now anyways.
Which if you really needed you could probably live with a USB 1.
Are those a SFF (2.5" in bay)
-
Yes
-
Depending on your IOPs, you could be looking at SSD or HDD storage.
@jasonlow said:
We probably don't NEED HA. We do have two decent servers so could probably drop down to one and use the second as a backup in case the first one goes down. I don't know the answer on the IOPs.
You think you can get away with one server for your entire workload? That definitely makes things easier.
You didn't mention what you are doing for backups?
You didn't mention a DB server other than Exchange, but you mentioned two EHR servers. Is one of them a DB server? If so, what DB? How many users on your EHR? RDS (terminal server is the old name)?
-
I would go with RAID 5 SSD's 1TB capacity.
The price per GB is more expensive but you'd be gaining all of the benefits of SSD storage.
You could go with RAID 10 2TB Drives but would need more disks to achieve your storage goals. Which might out weigh your cost-goal.
-
Moving from local disks to local USB and/or SD card is pretty easy since it is a fresh install combined with moving the settings over. ESXi is not stateless, but nearly so. So moving between media is not difficult.
-
@DustinB3403 said:
I would go with RAID 5 SSD's 1TB capacity.
The price per GB is more expensive but you'd be gaining all of the benefits of SSD storage.
He might not really gain any benefits if his IOPS are very low.
-
I'm starting from the top so might have missed details, but disk size is selected but not type, RAID level or spindle count? Those all have to be decided as a single, holistic decision based around speed, risk and cost needs.
What are the capacity, performance and cost needs for this?
-
@scottalanmiller said:
What are the capacity, performance and cost needs for this?
Yes, figure out your IOPS needs, business needs and budget before coming up with a solution. Don't make the solution before you've determined the needs.
-
@jasonlow said:
Is this a good idea? Pros and cons......Thanks!
Moving to local disks makes sense now that a purchase needs to be made. The SAN is negative in all four areas: cost, capacity, performance and risk. So adding to the technical debt there would be bad.
-
@scottalanmiller said:
I'm starting from the top so might have missed details, but disk size is selected but not type, RAID level or spindle count? Those all have to be decided as a single, holistic decision based around speed, risk and cost needs.
What are the capacity, performance and cost needs for this?
All things that DPACK and Danielle will help him with.
-
@jasonlow said:
Our SAN has 7.2TB running RAID 10, so approximately 3.6TB usable. There is about 415GB free. I like VMware and we own it so I hate to move away from it. We are a non-profit so our cost is feasible both on the VMware and Microsoft side.
It's not just cost but loss of features and extra license management work for you. VMware is not just a price negative.
-
@jasonlow said:
Currently there are 8 drive bays, but if I remove the optical drive I can expand to 16.
You can do a lot with eight bays. Sixteen really gives you flexibility.
-
@scottalanmiller said:
@jasonlow said:
Our SAN has 7.2TB running RAID 10, so approximately 3.6TB usable. There is about 415GB free. I like VMware and we own it so I hate to move away from it. We are a non-profit so our cost is feasible both on the VMware and Microsoft side.
It's not just cost but loss of features and extra license management work for you. VMware is not just a price negative.
Drat - I forgot to go back and mention this.
Why pay one penny if it gains you nearly nothing or nothing. Sure you're more familiar with it now, but really Hyper-V is pretty easy and straight forward... and offers no road blocks to features you might want to use in the future.
-
@jasonlow said:
We probably don't NEED HA.
You aren't even at "Standard Availability" today, you have low availability (quite a bit less reliable than just a single server) so if you need HA your current setup wouldn't have been an option. Even if you needed the reliability of a stand alone server what you have today is not there. So either you don't need either SA or HA, or your current situation is not meeting your needs and that needs to be addressed as well.
-
There are two decent choices going forward....
- Move to standard local storage. This will move you just from LA (low availability) to SA (standard availability) while being the cheapest solution. There are two ways to tackle this:
- Going to all local storage on a single server
- Going to half of the needed storage on each server and splitting load
- Move to replicated local storage. This will move you towards HA at the platform level (and only at that level, this doesn't actually give you HA unless you do everything else needed for HA too) but doubles the necessary investment in storage AND requires that you buy much more expensive licensing from VMware. The licensing would be free from any platform other than VMware.
- Move to standard local storage. This will move you just from LA (low availability) to SA (standard availability) while being the cheapest solution. There are two ways to tackle this:
-
While Scott likes to harp on people being below standard availability, I consider these statements less than helpful. While not true, I'm willing to bet that many organizations *believe * they have HA because of their SAN. So it's fine to tell them that their belief is wrong, but not that that wasn't their intention.
-
So what I'm hearing is local storage and running ESXi on a flash drive is the way to go. But I'm also hearing I should scrap VMware and move to Hyper-V?
We already have the VMware licensing for 2 hosts and according to VMware have HA so I'm not sure what the extra licensing costs would be if we stick with VMware.
We do have Unitrends if that changes anything with regards to replication.
-
@Dashrender said:
While Scott likes to harp on people being below standard availability, I consider these statements less than helpful.
I think it is extremely important to understand what was considered to be acceptable in the past when evaluating the future. You assume that people need HA and their mistake is having missed the mark in the past. Even if that is true, which presumably often it is, it is still incredibly important to know what they have today to know if they can keep it, leverage it, must replace it, etc.
In the other case, it may show that if what they have today is acceptable then they don't need HA no matter what they state in words. It just means that they do not know what HA is.
Under what possible condition would knowing the risk of the current scenario not be of the utmost criticality.
-
@Dashrender said:
. So it's fine to tell them that their belief is wrong, but not that that wasn't their intention.
Did I state that it was their intention? You are putting words in my mouth. People repeat that I say that so often that people believe that it is true. What I stated is that if what they have today is good enough, HA is not needed. I did not imply anything about their intentions. Very different things.
The assumption that I am being condescending is so strong that people hear it regardless of what I actually state. But I'm actually trying to provide unavoidably important reading of what knowledge we have.
If the current setup is "good enough" today, then we know something incredibly important. If it is not good enough, we know something else incredibly important.
-
For example: if the LA setup that exists today is not good enough in terms of availability, the it must be replaced on those grounds and cannot be maintained. This eliminates all LA options from our future solution set. So we cannot avoid this information without being forced to throw out the P2000 (from this particular system) in order to avoid an assumed need for HA.
But if we determine that HA and even SA are not needed and that the current system, which is LA, is truly "good enough" then we can use a hybrid approach going forward to leverage local storage and the P2000 to dramatically lower the cost of the future solution since, while we would not buy it new, it is perfectly acceptable to keep using it going forward.
Does that make it clear why this assessment is necessary for any discussion about future planning? We cannot just assume HA and the current solution suggests that HA is not needed either because that's not why it was purchased or because it has not been needed thus far or because HA is simply misunderstood and what they have is considered HA (misunderstanding HA in terminology but accepting the existing reliability as being above that which is necessary.)
