OpenIO
-
@scottalanmiller said in OpenIO:
I looked at the setup you did... So what kind of use cases would there be for this? It looks as if someone would have to build in an API or something to store objects in it?
S3 and Swift are built in. If you don't want to use those, then yes, you'd either need to build your own or buy one.
S3, I am familiar with. What is Swift?
Swift is the S3 equivalent from OpenStack. Same concept, different interface (SMB vs. NFS, potato po-tah-to)
-
So tools like CloudBerry or CyberDuck will talk to OpenIO no problem right out of the box.
-
@scottalanmiller said in OpenIO:
So tools like CloudBerry or CyberDuck will talk to OpenIO no problem right out of the box.
Ah, okay. That makes sense. My first thought jumped to "How could this be used to handle running things like VMs or file shares"... must be my brain locked in on current projects, lol.
-
@scottalanmiller said in OpenIO:
So tools like CloudBerry or CyberDuck will talk to OpenIO no problem right out of the box.
Ah, okay. That makes sense. My first thought jumped to "How could this be used to handle running things like VMs or file shares"... must be my brain locked in on current projects, lol.
Scale Out is really not a tool for VMs. It's possible, and that's what Swift itself does and Exablox can do that, but Scale Out design and object storage is rarely designed around low latency micro bursts like you want for a typical VM. It's designed for throughput performance and durability and, obviously, scaleability.
-
Just want to make sure everyone saw @scottalanmiller's how to on OpenIO: https://mangolassi.it/topic/10221/building-openio-on-centos-7
-
@Minion-Queen Yep! I saw. That's what made me ask about it. It looks interesting and all, but for those of us not used to working with Object based stuff, it'll take a while to get up to speed.
-
nice find ... I found some documentation on the OpenIO website itself .... Can't think of any usage scenario for us, or our clients... the minuscule quantum of data we handle does not warrant object based storage... but, very interesting product none-the-less ..
Thanks Scott
Hello @Veet,
Thanks for your interest!
OpenIO is an object storage software, and some of our users use it without storing a lot of data.
Why? Because it's also a new way to manage/deal with your storage. For example you can access to data through rest api, which could be very useful to access to it remotely (build your own S3 platform like mentioned by @scottalanmiller).
It's also a way to consolidate your storage usage, by using only one solution for many needs (storage as a service, archive, file sharing...) or many customers if you are a service provider. -
I take it at present time, that there's no way to "Mount" this like one would do with a file system?
-
@scottalanmiller said in OpenIO:
@scottalanmiller said in OpenIO:
So tools like CloudBerry or CyberDuck will talk to OpenIO no problem right out of the box.
Ah, okay. That makes sense. My first thought jumped to "How could this be used to handle running things like VMs or file shares"... must be my brain locked in on current projects, lol.
Scale Out is really not a tool for VMs. It's possible, and that's what Swift itself does and Exablox can do that, but Scale Out design and object storage is rarely designed around low latency micro bursts like you want for a typical VM. It's designed for throughput performance and durability and, obviously, scaleability.
Agree with you @scottalanmiller.
@dafyre to give you some examples of use cases suitable for object storage, some of our users built email platform, file sharing system, video streaming, storage as a service, archiving etc...
All these use cases have the same issue. They need performance, durability (replication of data to prevent from data loss) and scalability (meaning that being able to easily grow your platform, to follow your needs, but without any painful migration or task to perform. The scalability needs to be transparent for the application).
Object storage solutions bring that.
-
@GuillaumeDelaporte So essentially any application that wants to be built using OpenIO (or any other Object storage) has to be prepared to build based around the REST API?
-
-
@GuillaumeDelaporte So essentially any application that wants to be built using OpenIO (or any other Object storage) has to be prepared to build based around the REST API?
Yes, but applications and solutions which you can use are more and more compatible with this new style of storage (because it solves some problems you can face with other technologies).
-
I take it at present time, that there's no way to "Mount" this like one would do with a file system?
NFS and SMB are supported. iSCSI might be too, not sure. Those features are in the paid for Enterprise version.
-
@scottalanmiller said in OpenIO:
I take it at present time, that there's no way to "Mount" this like one would do with a file system?
NFS and SMB are supported. iSCSI might be too, not sure. Those features are in the paid for Enterprise version.
Correct. iSSCI is not supported ... yet.
-
@GuillaumeDelaporte said in OpenIO:
@GuillaumeDelaporte So essentially any application that wants to be built using OpenIO (or any other Object storage) has to be prepared to build based around the REST API?
Yes, but applications and solutions which you can use are more and more compatible with this new style of storage (because it solves some problems you can face with other technologies).
And, of course, if you are building your own.
MangoLassi stores certain things in this manner, for example.
