Power Loss Followthrough
-
Well, to take a step back, I always thought the reasoning behind enterprise class SSD was
a) -- power loss circuitry to prevent data loss
b -- longer write cycle life
c -- onboard cachingIf a RAID controller
a -- takes care of this for us
b -- absorbs all the writes
c -- disables this cachingWhy would one ever recommend an enterprise level SSD behind a raid controller? Or are you saying they (You) wouldn't?
EXCEPT if you needed to maintain system conformity (all DELL) or get a feature like the LEDs in DELL server that the EDGE SSDs give you?
-
@Dashrender said:
I don't think RAID controller vs non-RAID controller was his point.
Assumptions:
- All business class storage workloads are on RAID or equivalent (like RAIN).
- All business class RAID has cache.
-
@BRRABill said:
Well, to take a step back, I always thought the reasoning behind enterprise class SSD was
a) -- power loss circuitry to prevent data loss
b -- longer write cycle life
c -- onboard caching even in the Winchester era, this was disabled which is why enterprise drives had little to no cache and consumers have big cache. -
The reasons for enterprise SSDs are primarily:
- Integrated support from the server vendor
- Longer write lifetimes
-
@BRRABill said:
Why would one ever recommend an enterprise level SSD behind a raid controller? Or are you saying they (You) wouldn't?
With rare exception, only when you want integrated support from your vendor or need controller support for the firmware.
-
@scottalanmiller said:
- Longer write lifetimes
When would this be applicable, since you mention it here, but only recommend enterprise class in rare cases, none of which of the rare cases mention this.
-
@BRRABill said:
@scottalanmiller said:
- Longer write lifetimes
When would this be applicable, since you mention it here, but only recommend enterprise class in rare cases, none of which of the rare cases mention this.
When you have systems that are hard to get to, generally. Like on ships or in Antarctica and shipping replacement drives there is very costly and difficult.
-
So just to confirm
- if you aren't in Antarctica and
- you are assumed to be using a Raid card with cache, and
- you don't need manufacturer integrated support
...consumer is the way to go.
Regular consumer (like the Samsung EVO 850) or the Pro consumer (Evo 850 PRO)?
-
The problem that this boils down to is... you want to use Dell hardware, but non dell HDs. This is a mismatch.
@scottalanmiller , would you say that it's probably a good time for SMBs to consider moving away from Dell/HP and more toward SuperMicro because you can use your own drives more freely with the RAID controller you install there?
of course, you probably don't get things like front panel lights on failing/failed drives?
-
I'm not talking about my situation, just in general.
Concept and theory.
Just to learn.
I fully understand what I have done with the DELL drives.
-
I really originally never thought about why you;d have to worry about power loss if you have a UPS, which is what started this thread.
-
@BRRABill said:
I really originally never thought about why you;d have to worry about power loss if you have a UPS, which is what started this thread.
Because, if you're like me, it's definitely worth having the UPSs, but it's not worth it to install generators. Of course if you have short term battery power, say 15 mins, I suppose you should probably have scripts etc in place to start shutting down your servers when you loose power for more than say 5 mins.
-
@BRRABill said:
So just to confirm
- if you aren't in Antarctica and
- you are assumed to be using a Raid card with cache, and
- you don't need manufacturer integrated support
...consumer is the way to go.
Regular consumer (like the Samsung EVO 850) or the Pro consumer (Evo 850 PRO)?
Unless you want support. Support has always been the driver to enterprise gear. Whether servers, SAN, desktops, laptops, NAS, your name it... big business pays for vendor support. What types of gear they get are not things that they look at. It is only when you leave the enterprise world of "nothing but integrated support" do these questions even arise. Big businesses never part out their servers because integrated support is worth more than the gear itself.
-
@BRRABill said:
I really originally never thought about why you;d have to worry about power loss if you have a UPS, which is what started this thread.
If you use a UPS correctly, you really don't. That's why mainframes and mini computers (the $50K+ gear) don't have these kinds of protections at all. They expect you to be fully responsible for making sure that the power never fails and that if it is going to fail that you power the systems down properly first. The idea of baking the protection into the individual servers or controllers is an SMB approach where IT is trying to cover for facilities departments that are not being held accountable or can't be.
-
@Dashrender said:
@BRRABill said:
I really originally never thought about why you;d have to worry about power loss if you have a UPS, which is what started this thread.
Because, if you're like me, it's definitely worth having the UPSs, but it's not worth it to install generators. Of course if you have short term battery power, say 15 mins, I suppose you should probably have scripts etc in place to start shutting down your servers when you loose power for more than say 5 mins.
Yes, shutting down is always a viable option. You just need to make sure you have failover UPS that are properly monitored and maintained.
