Ask Your Question
0

Doesn't Swift use storage the same way Cinder does?

asked 2015-10-22 04:34:35 -0500

TimHas gravatar image

Hey guys,

I know Swift and Cinder store different kinds of data (Object vs Block), but I kind of imagened them both just using a bunch of HDDs/SSDs. I was looking at https://www.rdoproject.org/HA_Architecture (this RDO page) about High-Availability, and I noticed that there was a storage network with shared/distributed Storage. I thought this storage network would be for both Cinder and Swift, but the odd thing I noticed was that there were a couple of separate nodes for Swift, which where not connected to the Storage Network.

I'm currently working as an intern for this company and I'm trying to figure out how they could setup an OpenStack environment. The thing is that they have a SAN in their network and they want to use this for the storage of the Cinder and Swift services. So I kind of figured that we'd install Swift and Cinder on the same node and connect it to the management and storage network, and give it an iSCSI connection to the SAN.

This is why the separate Swift nodes, which are not connected to the storage network, threw me off. Is my idea for the setup not correct? Or is it just another way of doing things?

edit retag flag offensive close merge delete

Comments

Swift is designed to run on cheap servers with cheap directly attached JBOD or RAID. I don't think it is typically used with SAN storage. I suggest reading the white papers on http://swiftstack.com to familiarize yourself with this subject. On the other hand SAN is suitable for Cinder.

Bernd Bausch gravatar imageBernd Bausch ( 2015-10-23 23:49:28 -0500 )edit

1 answer

Sort by ยป oldest newest most voted
0

answered 2016-01-12 07:59:03 -0500

jdexter gravatar image

You are correct in that both types of storage are just made up of HDDs but the its how they present this storage. It is also how they are designed to retrieve this.
Swift is an object store, each piece of data is stored as an object, each object can be directly accessed via the API. Cinder is a block storage device, that data that cinder holds is part of a block device, the block device itself is connected to a client before it can be accessed.

Check out if you SAN is on the Cinder Matrix as if it is you can directly access the san via cinder.

image description

In the diagram above ScaleIO is taking the place of a SAN for Cinder to access, but as you can see the Swift storage nodes are on their own nodes, due to the resource requirements needed for them.

If you want to use your SAN resources for "swift" you would have to make a LUN on a server, then install the swift storage services on that server using the LUN to provide the storage.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2015-10-22 04:34:35 -0500

Seen: 447 times

Last updated: Jan 12 '16