SANs

[YoungGun21]

Can anybody explain to me what exactly a storage area network is and how it differs from network attached storage? I'm not really getting the difference.

 

[theevilsharpie]

Network-attached storage devices expose their storage as a networked file system. SANs expose their storage as a block device.

Using Windows as an example, you would use a NAS by accessing a shared volume or mapping a network drive. You would use a SAN the same way that you would use an internal disk drive (e.g. partitioning, formatting with a file system, etc.).

This is a very simplified explanation, but I hope it gets the basic difference across.

 

[YoungGun21]

Ok once I looked up block devices this makes a little bit more sense.

Could you explain essentially like how it works? What is used? Servers?

If I'm using a computer in a computer lab at my school, can I tell a difference between the school using NAS or SAN? Can the computer tell the difference?

 

[theevilsharpie]

Explaining how SAN's and NAS's work is a bit involved, and I would suggest looking up the relevant Wikipedia articles.

Pretty much anything that can share storage using a networked file system would be considered a NAS. If you're sharing any files from your computer, you've turned it into a NAS. NAS's typically use SMB, NFS, or FTP as the NAS protocol.

SAN's are usually purpose-built units with software that are designed specifically for SAN use. Most SAN's are provided by commercial vendors such as HP, Hitachi, EMC, etc., but free software packages exist that allow you to set up an lightweight SAN. SAN's typically use Fibre Channel or iSCSI as the SAN protocol.

I've already described the technical differences between SAN's and NAS's, so here's two articles that walk you through the process of connecting to a network share (NAS) and an iSCSI (SAN) from Windows 7, which should better illustrate the difference between using SAN's and NAS's.

NAS:
http://www.7tutorials.com/how-map-drives-windows-7

SAN:
http://www.windowsnetworking.com/articles_tutorials/Connecting-Windows-7-iSCSI-SAN.html

 

[YoungGun21]

Yea I read the wiki articles, but I couldn't tell the difference just from those. I'll read through those links and try to find more articles online.

Thanks

 

[Mark R]

The key point, as has already been said, is that a SAN appears as a "block device" (a hard drive) to the OS. As far as the OS can tell, the SAN is a hard drive which is directly connected to the computer. If you have a dedicated SAN card in the computer (or use a high-end LAN card with hardware SAN support), then you can even do stuff like have no drives in the computer - just install the OS and boot from a SAN drive.

A SAN (because each "partition" on it acts like a directly connected hard drive) can only have each "partition" accessed by 1 computer at a time. Additionally, because the SAN acts like a drive, the comp will have full access to that "drive".

E.g. I have a 1 TB SAN, and I create 4 250 GB "drives". Comp A, can connected in to drive A. Comp B can connect in to drive B, etc. and each comp will see the "drive" appear as a 250 GB hard drive (which then needs partitioning, formatting, etc.). However, 2 comps cannot connect to 1 "drive" simultaneously (in the same way that you can't connect a SATA drive to 2 comps simultaneously).

A SAN is most useful for working with multiple servers. All the servers use disk space on the SAN rather than their own hard drives. Let's say server X blows a CPU. With SAN, you just go to a spare server, configure the SAN card to connect to "drive X" on the SAN, boot it up, and you're done. You've effectively transplanted the hard drives from X into the spare - and you can do it all remotely.

A NAS appears to the OS as a "shared drive" - a collection of files which can be accessed by multiple computers. Multiple comps can access a single file simultaneously (as long as none of them changes the file).

A NAS is most useful if you want multiple different files to be accessible from multiple places, or if you want to restrict access to individual files by user.

 

[her209]

A SAN (because each "partition" on it acts like a directly connected hard drive) can only have each "partition" accessed by 1 computer at a time. Additionally, because the SAN acts like a drive, the comp will have full access to that "drive".

I think its dependent on the file system. NTFS can only be access by one system at a time but VMFS can be accessed by multiple systems.

 

[Emulex]

CSV can be accessed by multiple machines on windows. (clustered storage volume) same as vmfs is for vmware


CSV and VMFS allow the hypervisor of each o/s to have more than 1 computer access the volume of vm(s). for fault tolerance and for backup or hot migration(vmotion).

 

[Nothinman]

I think its dependent on the file system. NTFS can only be access by one system at a time but VMFS can be accessed by multiple systems.

Yes, most SANs let you specify whether an iSCSI target can be accessed by more than one system at a time but whether it really works or not depends on the hosts involved. There are a number of cluster filesystems around like GFS, OCFS, VMFS, etc but I think all of the ones for Windows are commercial and probably quite expensive.

 

[YoungGun21]

So in the case of like a University or large company, each person gets a personal folder which is really just a partition of a HDD in the server room? So when I log in using my credentials it just gives me access to that particular partition?

Also, I'm guessing SANs are faster?

 

[theevilsharpie]

So in the case of like a University or large company, each person gets a personal folder which is really just a partition of a HDD in the server room? So when I log in using my credentials it just gives me access to that particular partition?

End-users are rarely given direct access to storage on a SAN. Storage on a SAN is usually provided to an organization's servers, and if necessary, these server might expose this storage to end users using a networked file system.

If you're logging in to a server using Citrix, Remote Desktop, or some sort of virtual desktop, there's a strong possibility that the storage being used is on a SAN.

Also, I'm guessing SANs are faster?

In theory, there isn't a performance difference between a SAN and a NAS, although it's a bit difficult to compare them directly because they do different things. In practice, SAN's are usually much faster, simply because their use case (back-end server storage) demands more performance than a typical NAS use case (bulk file storage) and the systems are larger as a result.

 

[YoungGun21]

The key point, as has already been said, is that a SAN appears as a "block device" (a hard drive) to the OS. As far as the OS can tell, the SAN is a hard drive which is directly connected to the computer. If you have a dedicated SAN card in the computer (or use a high-end LAN card with hardware SAN support), then you can even do stuff like have no drives in the computer - just install the OS and boot from a SAN drive.

A SAN (because each "partition" on it acts like a directly connected hard drive) can only have each "partition" accessed by 1 computer at a time. Additionally, because the SAN acts like a drive, the comp will have full access to that "drive".

E.g. I have a 1 TB SAN, and I create 4 250 GB "drives". Comp A, can connected in to drive A. Comp B can connect in to drive B, etc. and each comp will see the "drive" appear as a 250 GB hard drive (which then needs partitioning, formatting, etc.). However, 2 comps cannot connect to 1 "drive" simultaneously (in the same way that you can't connect a SATA drive to 2 comps simultaneously).

A SAN is most useful for working with multiple servers. All the servers use disk space on the SAN rather than their own hard drives. Let's say server X blows a CPU. With SAN, you just go to a spare server, configure the SAN card to connect to "drive X" on the SAN, boot it up, and you're done. You've effectively transplanted the hard drives from X into the spare - and you can do it all remotely.

A NAS appears to the OS as a "shared drive" - a collection of files which can be accessed by multiple computers. Multiple comps can access a single file simultaneously (as long as none of them changes the file).

A NAS is most useful if you want multiple different files to be accessible from multiple places, or if you want to restrict access to individual files by user.

I thought that is what the bolded is saying here.

I am teh suck at networking.

 

[mcturkey]

I thought that is what the bolded is saying here.

I am teh suck at networking.

Think of a SAN as one really big computer made up of several smaller computers. Instead of a SATA cable running between the controller card ("Head Node") and the hard drive ("Storage Node"), it's all interlinked via network cables. This allows for massive scalability, while providing greater redundancy options.

If you need, say, 1PB of storage space directly accessible from a single location, there is no way to fit the hundreds of disks required into a single chassis. In order to accomplish this, you'd have to either daisy chain SAS connections and RAID cards to create a really convoluted mess of cases and cables that could never be more than a couple feet apart and would be a maintenance nightmare, or use a SAN. Then you can have 20 4U boxes filled with hard drives, all acting as an extension of a head node ("system"), which would present the storage to whatever server functionality needs it. That head node then handles RAID, backup, drive management, etc as if it had 500 disks directly attached to the motherboard or a RAID card.

 

[Red Squirrel]

Often a SAN is actually exactly that, a storage area network. It will have a dedicated switch or vlan, but usually switch. The servers will have dedicated nics to connect to that network. The SAN will present "shares" to the servers, often through iSCSI, but sometimes NFS. There is also something called a HBA (host bus adapter) that is used instead of a nic. Basically the HBA will talk to the SAN and act at the hardware level. It will present the volumes to the computer. This is common with fiber channel but can also be done with iSCSI. iSCSI is basically a protocol to share volumes to an OS or computer and that computer will present them as a locally attached device. So say you make a raid 5 array on a SAN and share it as an iSCSI volume, you initialize it on the server and you'll see a new disk in device manager. It's a standard so it works in Linux too. You have to format it on the server itself, the SAN just sees it as block level, not file system level.

SANs are typical in VM environments. All the VMs are stored on it, and the VM servers spin off the VMs from the SAN. Those VMS are then presented to the real network. One advantage of a setup like this is that the data is central and the servers more or less act as appliances. You could have a spare server sitting somewhere, if one craps out, you just put in the spare server, no need to migrate data or do anything. Often high end SANs will have redundancy and high availability features too. So your data is in one basket, but that basket is very strong.