Why Choose AetherStore Over a NAS?
-
@stacksofplates said in Why Choose AetherStore Over a NAS?:
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@stacksofplates said in Why Choose AetherStore Over a NAS?:
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@aidan_walsh There are a bunch of DFS's in existence, but are you really capable of deploying them to your working fleet of windows PC's?
One of the key things with AetherStore is you're supported, and the cost is negligible compared to the weeks or months of setting up your own implementation.
Seems crazy that list doesn't have GFS2.
GFS2 is a CFS, not a DFS. It goes on a SAN.
I thought you could do parallel read/write over multiple nodes with it also.
You can, that's a CFS. But it's just a filesystem. DFS handle the distribution, CFS just work when the underlying system is distributed OR when it is not. For example, CFS work on a SAN, DFS replace a SAN.
-
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@stacksofplates said in Why Choose AetherStore Over a NAS?:
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@stacksofplates yeah... idk why it's not there. Is GFS2 only available on RedHat?
Seems odd that it would be....
I mean you might be able to use it on something else, I never looked into it but it's developed by RedHat.
Developed by RedHat doesn't mean it's available publicly, like RedHat isn't available publicly under the RedHat name. It's available under the CentOS name.
So maybe it's named something different on CentOS?
Actually it does I think, but only because of corporate policy.
-
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@stacksofplates said in Why Choose AetherStore Over a NAS?:
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@stacksofplates said in Why Choose AetherStore Over a NAS?:
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@aidan_walsh There are a bunch of DFS's in existence, but are you really capable of deploying them to your working fleet of windows PC's?
One of the key things with AetherStore is you're supported, and the cost is negligible compared to the weeks or months of setting up your own implementation.
Seems crazy that list doesn't have GFS2.
GFS2 is a CFS, not a DFS. It goes on a SAN.
I thought you could do parallel read/write over multiple nodes with it also.
You can, that's a CFS. But it's just a filesystem. DFS handle the distribution, CFS just work when the underlying system is distributed OR when it is not. For example, CFS work on a SAN, DFS replace a SAN.
Ah ic what you mean.
-
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@stacksofplates said in Why Choose AetherStore Over a NAS?:
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@stacksofplates yeah... idk why it's not there. Is GFS2 only available on RedHat?
Seems odd that it would be....
I mean you might be able to use it on something else, I never looked into it but it's developed by RedHat.
Developed by RedHat doesn't mean it's available publicly, like RedHat isn't available publicly under the RedHat name. It's available under the CentOS name.
So maybe it's named something different on CentOS?
Actually it does I think, but only because of corporate policy.
I think it's just gfs2-utils
-
Ubuntu 16.10 has GFS2
apt-get install gfs2-utilsJust tested on my laptop.
-
@stacksofplates said in Why Choose AetherStore Over a NAS?:
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@stacksofplates said in Why Choose AetherStore Over a NAS?:
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@stacksofplates yeah... idk why it's not there. Is GFS2 only available on RedHat?
Seems odd that it would be....
I mean you might be able to use it on something else, I never looked into it but it's developed by RedHat.
Developed by RedHat doesn't mean it's available publicly, like RedHat isn't available publicly under the RedHat name. It's available under the CentOS name.
So maybe it's named something different on CentOS?
Actually it does I think, but only because of corporate policy.
I think it's just gfs2-utils
Yes, that's its name.
-
Here is a guide to using it.
https://www.server-world.info/en/note?os=CentOS_7&p=pacemaker&f=3
-
Something I want to see from AetherStore is a multi-node exposure option so that we can put GFS2 or OCFS on it.
-
Hey Everyone, tons of great questions on AetherStore here! We’ve been enjoying watching the thread keep expanding over the past few days and have clarified some important points below. Please let us know if there are any other questions, we’re happy to expand on anything you want to know more about.
Redundancy: 4x is the default but this is customizable. We price per usable (fully replicated) space, so if you believe you need increased availability, the pricing model rewards higher replication. AetherStore also proactively maintains your redundancy level. As machines go down it will recopy chunks to live nodes to maintain full replication. Also important: machines going offline does not cause data loss. If you need to read or write to the drive, yes - some machines need to be turned on.
As @scottalanmiller pointed out, AetherStore will not allow you to write to the drive if it cannot achieve full replication. In past versions a write could be completed locally even if not enough machines were online, assuming redundancy would be achieved when machines come back on. We’ve changed the way writes are completed in AS 2.0 to prioritize redundancy instead. So now, every time you’ve completed a write you know it’s been successfully written to (n) machines. In the future, we look forward to introducing write profiles, so you can select how you would prefer to write.
Speed: AetherStore is a robust, affordable and secure backup target that provides quick enough read times to serve its intended purpose well – reducing downtime, helping you meet RTO when the crunch is on. That said, if fast write speeds are your most important requirement AetherStore may not be the product for you right now. For every write, data is chunked, encrypted (twice) and stored across many machines. The same is true for the metadata, and there are environmental factors in every network to consider. These things take time. AetherStore’s complexity under the hood is what enables it to be a truly resilient, secure, flexible product.
Resource Consumption on Machines: @NetworkNerd: what @JaredBusch said - nothing noticeable. We run AetherStore constantly and have not observed effects on the machines nor has it been reported by users. We do have numbers on this, our QA engineer is currently updating them for the 2.0 release so I will send them through when I have them!
Mount Machine: What you guys are referring to as the “head node” here. You’re correct that AetherStore supports one mount machine at a time. You can change the mount machine after deploying a Store at any time using the dashboard. If the mount machine goes down, just use the Dashboard to select a new one. The mounted drive is locked and password protected. It can also be unmounted completely which provides a cool veiled mode for the drive. Data remains perfectly intact but the drive is invisible/inaccessible until remounted. As @scottalanmiller pointed out, you don’t need the mount machine to contribute a certain amount of storage. However, choosing a fast/good machine as the mount will definitely help performance!
Linux: we don’t support Linux in the current release but it’s far not off. I’ll keep you updated on it!
AetherStore 2.0 Release: we’re in the process of rolling this out now! If you gave us your info at MangoCon you will already be in one of the Early Release groups and should receive an email shortly.
Thanks again for all the great input!
-Shannon
-
@shannon Thanks!!
-
@shannon This is a generic question that could vary a lot with the answer.
How many nodes would have to be lost to be unable to restore?
I know I know, its a very generic question.
-
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@shannon This is a generic question that could vary a lot with the answer.
How many nodes would have to be lost to be unable to restore?
All of them
-
There is no simple answer for you Dustin. If you have four nodes and four times replication, then you'd have to lose every single one to not be able to restore. Any one would contain all of the data.
-
@scottalanmiller Really, how does ever node contain all of the data?
If you have a 500GB hard drive, with 300GB free, it realistically can only store 300GB of data. Edit: Not an entire Mirror copy of the data 1/4th. It must hold something less than that.
So my question is, at what kind of disaster has to occur to cause AetherStore to be unable to restore from?
Maybe I'm missing something. @Rob should come answer this
Also @Rob are you coming to MangoCon this year?
-
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
@scottalanmiller Really, how does ever node contain all of the data?
If you have RAID 1 on four drives, each drive contains all of the data. You have to lose all four drives to lose anything.
This isn't RAID 1, but it is RAIN acting like RAID 1 with four time mirroring in a case of four nodes with the stock four times replication.
-
@DustinB3403 said in Why Choose AetherStore Over a NAS?:
If you have a 500GB hard drive, with 300GB free, it realistically can only store 300GB of data?
Yes, a drive with only 300GB free can only store 300GB of data. That's not really an AetherStore question
-
Awesome post @shannon thanks for giving us the lay of the land.
-
Thanks @shannon
-
@shannon said
Speed: AetherStore is a robust, affordable and secure backup target that provides quick enough read times to serve its intended purpose well – reducing downtime, helping you meet RTO when the crunch is on.
Go on then, tell us the figures
-
Lots of factors there, I'm sure. The biggest question would be, I supposed, how fast would AetherStore be if only it were the bottleneck.
