Why Choose AetherStore Over a NAS?
-
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@Dashrender said in Why Choose AetherStore Over a NAS?:
My machines mostly still have HDDs in them. 500 GB drives. Probably 300+ GB free. If I ignore the laptops (yes I know Laptops can be part of it.. AetherStore specifically tested their coming and going as part of their tests), I have around 50 machines. Assuming only doubling of the data, that would provide me with 7.5 TB of space total.
Now the question is, what's the performance like?
A lot of the performance comes from the main access node(s), so if those have fast drives it can make a big difference, or should.
So is the proposal that the main access node have somewhere around 50% or more of the required space of the entire array so that hopefully most of the data is coming from a single point?
OK I just pulled that out of my rear.. but if not for some huge amount of cache on the main node, I'm not sure what you're driving at?
-
@Dashrender said in Why Choose AetherStore Over a NAS?:
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@Dashrender said in Why Choose AetherStore Over a NAS?:
My machines mostly still have HDDs in them. 500 GB drives. Probably 300+ GB free. If I ignore the laptops (yes I know Laptops can be part of it.. AetherStore specifically tested their coming and going as part of their tests), I have around 50 machines. Assuming only doubling of the data, that would provide me with 7.5 TB of space total.
Now the question is, what's the performance like?
A lot of the performance comes from the main access node(s), so if those have fast drives it can make a big difference, or should.
So is the proposal that the main access node have somewhere around 50% or more of the required space of the entire array so that hopefully most of the data is coming from a single point?
If you want to treat the head node as a cache, yes, you could load it with fast storage, even give it SSD and even local RAID if you want. Depends on your goals. It's just one part of the puzzle so a question like this needs a lot of context.
-
@Dashrender said in Why Choose AetherStore Over a NAS?:
OK I just pulled that out of my rear.. but if not for some huge amount of cache on the main node, I'm not sure what you're driving at?
If you have a four node system and four time replication, that means that everything is on the head node - it won't go out to other nodes for other reads at all. If you have four time replication on eight nodes, half of all reads (all things being equal) half of your reads will come from the head node. No matter what, it's significant.
-
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@Dashrender said in Why Choose AetherStore Over a NAS?:
OK I just pulled that out of my rear.. but if not for some huge amount of cache on the main node, I'm not sure what you're driving at?
If you have a four node system and four time replication, that means that everything is on the head node - it won't go out to other nodes for other reads at all. If you have four time replication on eight nodes, half of all reads (all things being equal) half of your reads will come from the head node. No matter what, it's significant.
Sure, but these numbers seem pretty low. But maybe I'm missing something. I would anticipate that that those using something like this would have more like 20+ workstations. But of course, it is just as usable at the 4/8 node count as well.
But looking at the 20+ node (40ish in my case) now you're head node will have only a tiny fraction of the data on it.
-
@Dashrender said in Why Choose AetherStore Over a NAS?:
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@Dashrender said in Why Choose AetherStore Over a NAS?:
OK I just pulled that out of my rear.. but if not for some huge amount of cache on the main node, I'm not sure what you're driving at?
If you have a four node system and four time replication, that means that everything is on the head node - it won't go out to other nodes for other reads at all. If you have four time replication on eight nodes, half of all reads (all things being equal) half of your reads will come from the head node. No matter what, it's significant.
Sure, but these numbers seem pretty low. But maybe I'm missing something. I would anticipate that that those using something like this would have more like 20+ workstations. But of course, it is just as usable at the 4/8 node count as well.
But looking at the 20+ node (40ish in my case) now you're head node will have only a tiny fraction of the data on it.
Overall I think that that is exceedingly uncommon. Why would you choose the head node to be one with so little free storage, for one thing? And how commonly do people need twenty or more workstations to get the space that they need? Most have HDs and are getting loads of space per machine, not small SSDs. The small space per machine issue is real, but not, I believe, the majority case. And once you get to any size, the purchase of a single hard drive to significantly speed up operations is nothing.
Basically, if you are so small that a drive is an expense at all, it's probably not a factor to you. If you are big enough to care, a single drive is essentially free.
-
@Dashrender said in Why Choose AetherStore Over a NAS?:
But maybe I'm missing something. I would anticipate that that those using something like this would have more like 20+ workstations.
I don't know, but my impression is that what people tend to use is much smaller on average. But the bigger the pool, the lower the impact of individual nodes, too.
-
I think the biggest point I see is around security. With a backup server or multiple backup servers the data is contained to those machines wherever you decide to put them. It seems like going backwards having data shared locally on workstations. We make it a point to have nothing on the local workstations except what is needed for authentication and mounting of the real stuff.
-
@stacksofplates said in Why Choose AetherStore Over a NAS?:
I think the biggest point I see is around security. With a backup server or multiple backup servers the data is contained to those machines wherever you decide to put them. It seems like going backwards having data shared locally on workstations. We make it a point to have nothing on the local workstations except what is needed for authentication and mounting of the real stuff.
Heavily encrypted on the workstations. The individual workstations have no access to the data. And if you have more machines than you have redundancy, the data on the individual machines are incomplete, even when encrypted.
-
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@stacksofplates said in Why Choose AetherStore Over a NAS?:
I think the biggest point I see is around security. With a backup server or multiple backup servers the data is contained to those machines wherever you decide to put them. It seems like going backwards having data shared locally on workstations. We make it a point to have nothing on the local workstations except what is needed for authentication and mounting of the real stuff.
Heavily encrypted on the workstations. The individual workstations have no access to the data. And if you have more machines than you have redundancy, the data on the individual machines are incomplete, even when encrypted.
I guess my point is, if the right machines are stolen (since they are likely to have the least security), there is no backup.
-
@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?:
I think the biggest point I see is around security. With a backup server or multiple backup servers the data is contained to those machines wherever you decide to put them. It seems like going backwards having data shared locally on workstations. We make it a point to have nothing on the local workstations except what is needed for authentication and mounting of the real stuff.
Heavily encrypted on the workstations. The individual workstations have no access to the data. And if you have more machines than you have redundancy, the data on the individual machines are incomplete, even when encrypted.
I guess my point is, if the right machines are stolen (since they are likely to have the least security), there is no backup.
True, but it would take multiple, low value machines being taken to do that whereas a NAS is a single high value device. Which is more likely to be stolen? I don't really know. Depends on the environment. I've known places to have things like servers stolen, but not desktops. Laptops yes, but not desktops. Typically desktops are not worth it to fence. And you'd need to get all of the copies of the backups, which are potentially spread out to different rooms, offices, or whatever. Clearly something to consider, though.
But a single NAS or pair of NAS would be easily identifiable as things to steal or damage. I'd guess, even with more security, that centralized backups are more at risk in average situations.
-
If you have an office with just four desktops, and you use all four for AetherStore, and they all sit next to each other in a common room. Yup, that's likely some bad security of the backups.
-
@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?:
I think the biggest point I see is around security. With a backup server or multiple backup servers the data is contained to those machines wherever you decide to put them. It seems like going backwards having data shared locally on workstations. We make it a point to have nothing on the local workstations except what is needed for authentication and mounting of the real stuff.
Heavily encrypted on the workstations. The individual workstations have no access to the data. And if you have more machines than you have redundancy, the data on the individual machines are incomplete, even when encrypted.
I guess my point is, if the right machines are stolen (since they are likely to have the least security), there is no backup.
True, but it would take multiple, low value machines being taken to do that whereas a NAS is a single high value device. Which is more likely to be stolen? I don't really know. Depends on the environment. I've known places to have things like servers stolen, but not desktops. Laptops yes, but not desktops. Typically desktops are not worth it to fence. And you'd need to get all of the copies of the backups, which are potentially spread out to different rooms, offices, or whatever. Clearly something to consider, though.
But a single NAS or pair of NAS would be easily identifiable as things to steal or damage. I'd guess, even with more security, that centralized backups are more at risk in average situations.
Right but with the server, it's usually in a secured room or not even on site, where as the actual desktops/laptops aren't usually secured and are out in the open (depending on what type of office).
-
But I'll take an example client... they have hundreds of machines spread out all over the building. Some are in the warehouse - high risk, skip those. They are also old and on slow networks. But loads of executives, lawyers, accountants, engineers and such that have nice, newer workstations with big disks that are in individually locked offices. Those are low risk and the desktops are very separate from each other as risks go.
-
Right I mean we can hypothesize about anything. My point was just that's a factor that needs to be taken into consideration. Esp if you have different types of data replicated in differently (if you can do that).
To me it seems like a niche market. A place that's small enough to not have a real NAS (appliance or built) or tape, but not using SaaS. If you're big enough to have a NAS/tape (and real offsite backups), I don't see how this is a benefit.
-
Also, not that this isn't a cool product. It's really interesting.
-
@stacksofplates said in Why Choose AetherStore Over a NAS?:
Right I mean we can hypothesize about anything. My point was just that's a factor that needs to be taken into consideration. Esp if you have different types of data replicated in differently (if you can do that).
You can, if you split them into different pools.
-
@stacksofplates said in Why Choose AetherStore Over a NAS?:
If you're big enough to have a NAS/tape (and real offsite backups), I don't see how this is a benefit.
Because you have to be big enough for it to be a tape library with a robot or tape remains incredibly cumbersome. And the point here was that it many cases it might be superior to a NAS. So big enough to have a NAS and "should chose a NAS over it" are different things. I was asked why someone would chose AetherStore over a NAS for a small office, so I was demonstrating.
Is it for everyone? Of course not, nothing is. But I think that it has really solid use cases. Lots of shops really do have loads of excess hard drive capacity sitting around the building that can be used for loads of archival, backup or other capacity for very cheap.
-
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@Dashrender said in Why Choose AetherStore Over a NAS?:
@scottalanmiller said in Why Choose AetherStore Over a NAS?:
@Dashrender said in Why Choose AetherStore Over a NAS?:
OK I just pulled that out of my rear.. but if not for some huge amount of cache on the main node, I'm not sure what you're driving at?
If you have a four node system and four time replication, that means that everything is on the head node - it won't go out to other nodes for other reads at all. If you have four time replication on eight nodes, half of all reads (all things being equal) half of your reads will come from the head node. No matter what, it's significant.
Sure, but these numbers seem pretty low. But maybe I'm missing something. I would anticipate that that those using something like this would have more like 20+ workstations. But of course, it is just as usable at the 4/8 node count as well.
But looking at the 20+ node (40ish in my case) now you're head node will have only a tiny fraction of the data on it.
Overall I think that that is exceedingly uncommon. Why would you choose the head node to be one with so little free storage, for one thing?
Why would you think this was uncommon? The biggest initial issue in the first alpha/beta cycle was dealing with a VM head node with no storage at all beyond the OS partition. I am not going to buy or dedicate a machine with a large drive as a head node just to get increased performances. That is not the promise of this product.
After that, it was the horrible speeds.
I still have not used it again because the Alpha/Beta promised at MangoCon has never emerged.
I love the idea of this product as a way to not buy a NAS for backups.
But I cannot use something that does not work.
-
@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?:
I think the biggest point I see is around security. With a backup server or multiple backup servers the data is contained to those machines wherever you decide to put them. It seems like going backwards having data shared locally on workstations. We make it a point to have nothing on the local workstations except what is needed for authentication and mounting of the real stuff.
Heavily encrypted on the workstations. The individual workstations have no access to the data. And if you have more machines than you have redundancy, the data on the individual machines are incomplete, even when encrypted.
I guess my point is, if the right machines are stolen (since they are likely to have the least security), there is no backup.
That was part of my original question, in that there is no remote backup.
@scottalanmiller mentioend rmeote stores, but don't they have to be on the same LAN?
-
@BRRABill 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?:
I think the biggest point I see is around security. With a backup server or multiple backup servers the data is contained to those machines wherever you decide to put them. It seems like going backwards having data shared locally on workstations. We make it a point to have nothing on the local workstations except what is needed for authentication and mounting of the real stuff.
Heavily encrypted on the workstations. The individual workstations have no access to the data. And if you have more machines than you have redundancy, the data on the individual machines are incomplete, even when encrypted.
I guess my point is, if the right machines are stolen (since they are likely to have the least security), there is no backup.
That was part of my original question, in that there is no remote backup.
@scottalanmiller mentioend rmeote stores, but don't they have to be on the same LAN?
Does LAN over VPN count?
