Aetherstore in the real world
-
Yes this is what I'm trying to get at. Rule out what the product definitely is and what it definitely is not.
@scottalanmiller said:
Shut it down? Do you mean the node that has the DAS connection access?
Yes, the node which has mount access, which could be a DC or file-server.
-
@Breffni-Potter said:
@dafyre said:
@Breffni-Potter said:
@scottalanmiller said:
@Breffni-Potter said:
- How does it protect against cryptolocker and the like? How do you know which nodes are "clean" and which are corrupt?
No, it is a DAS. You need to provide the protection BEFORE you let things access the block storage. This acts just like any SAN would in this scenario.
But assuming the protection fails (and sometimes they do) - Is the entire pot of data lost in that scenario?
Not permanently. The systems all come back online and resync and figure out who has the latest "good" data.
So let's assume, Crypto has struck, found the endpoint which had it and shut it down, do the nodes automatically run through that process?
They would, yes. Aetherstore can sustain the loss of a node in various situations and not have anything affected. However, if crypto locker hits the data stores in Aetherstore... Have fun restoring from backups, lol.
-
@Breffni-Potter said:
Yes this is what I'm trying to get at. Rule out what the product definitely is and what it definitely is not.
@scottalanmiller said:
Shut it down? Do you mean the node that has the DAS connection access?
Yes, the node which has mount access, which could be a DC or file-server.
It if goes offline, the Cryptolocker would have no path to the storage.
-
@scottalanmiller said:
It if goes offline, the Cryptolocker would have no path to the storage.
...yes?
-
But you wouldn't know to offline that node or else you could have stopped CL anyway, one would assume.
Generally you want to protect any SAN or DAS connection point pretty heavily.
-
@scottalanmiller said:
But you wouldn't know to offline that node or else you could have stopped CL anyway, one would assume.
Generally you want to protect any SAN or DAS connection point pretty heavily.Indeed but important to clarify.
-
Hey guys - how goes it? Just popped by to say hi but it looks like I can help with a Q too: @Breffni-Potter if you have a 10-node Store, for example, and one of the nodes is hit by CL, you will still have access to your data, period. If the -mount node- is hit by CL however, and CL maliciously encrypts all drives it can write to (so including any mapped network drives etc.) then it would maliciously encrypt the Store data too.
In general, you can put a bullet in any machine running AetherStore and not notice. AetherStore will notice and automatically re-replicate the data that was on that machine onto other nodes in the system - no manual intervention required and your data is still accessible while this goes on in the background.
-
*I guess you'd notice if it was the machine you were working on at the time
plus, the noise... -
@Rob said:
In general, you can put a bullet in any machine running AetherStore and not notice.
This is what appeals the most.
I've got 6 Windows 7 test nodes and 1 2012 server live at the moment. I'm deliberately trying to break it but so far it keeps coming back to life.
-
@Breffni-Potter said:
I'm deliberately trying to break it but so far it keeps coming back to life.
That's what we like to hear
very good. -
@Breffni-Potter said:
- How would we monitor Aetherstore automatically? What if nodes fail/go down? How do you find out? Does it store in windows event logs?
- I cannot use Aetherstore dashboard without a license but if I install it onto a different machine with a license, I can immediately manage my existing stores on the network, including choosing a new mount point, am I only allowed to touch stores with the right license key? Or are all stores open to any dashboard manager?
Just in case these get lost.
-
I think I broke it.
Apart from choosing another node to mount, what can I do to get the client which is supposed to be mounted to show the store drive, computer management/disk management is empty. Is there a remount command?
-
1GB to 1.5GB of memory usage when using the dashboard.
What does it use in memory? Does memory usage increase with more nodes?
-
@Breffni-Potter looks like you killed the AetherStore Drive Manager process, or the AetherStore Daemon service. Restarting the Daemon service or running 'aetherstore-drive-manager.exe' in Program Files -> AetherStore -> Core will cause the drive to re-mount on that machine. If you killed the process manually then you probably know how to restart it
an alternative (and much simpler) solution is to just log out/log in and it'll re-mount by itself. -
@Rob said:
@Breffni-Potter looks like you killed the AetherStore Drive Manager process, or the AetherStore Daemon service.
Not deliberately
-
Is this a glitch? Or a sign of something wrong with the data.
-
That's a known timezone translation glitch in that spot on UI - great catch!
-
@Breffni-Potter ah - well-spotted. Please refer to the 'last sync' column in the data grid for the time being - looks like we've addressed that in v1.2.4 (releasing soon). As an FYI: you'll get a notification on the home screen of your dashboard when there's a new version available, and there's an 'update all' button that will push out the new version to your nodes automatically.
I see a backlog of a couple q's - gotta finish up some work here at the moment but I'll hop back in a bit and make sure to get you across all the info you need
-
Some performance tests
When Aetherstore is first powered up, the average speed is around 10 MB per second.
After it is settled, it can go as high as 20 MB per second but the average is around 15.
The read speed can go as high as 25 MB per second. -
The test network are virtual machines running on SSDs, the VMs themselves measure performance of 450 MB per second when testing the
drive. This performance was measured on the node where the drive was mounted.http://mangolassi.it/topic/7020/ntg-partners-with-aetherstore/30
So the performance matches what Aetherstore says it will be.
After dumping a load of data into a store, the read/write speeds do not suffer whilst it does the background task of replicating to other nodes. Even if I force a sync from the dashboard.
So far so good in terms of performance.
-
Awesome, thanks for the lab work!
