1

Possible Duplicate:
What is the use-case for NAS on a mid-to-high end storage array?

SAN is block level, NAS is file level. Block level means applications access blocks,while in file level they use files.

Now, we can also create file system on server operating system for block level storage, so what is the advantage of using file level access with NAS.

Please clear my doubts and mistakes.

Community
  • 1
Drt
  • 394
  • 2
  • 5
  • 18

4 Answers4

3

When using SAN the OS and Filesystem needs to be aware that changes to the blocks can be made by several systems at one, creating possible problems on the filesystem.

Some systems can handle that (designed for this type of access) like VMware vmfs, but most (both server and client) systems are designed to be singe-access to the file-systems on a block-level => if there is only a slight chance of duplicate access you would really want to use some file-level access, which can be handled in these systems ( = multiuser) => NAS

tsg

Truebsalgeblaese
  • 364
  • 1
  • 4
  • That sounds correct. Which filesystems would be suitable or works in this scenario? – Drt Jan 16 '13 at 10:26
  • 1
    VERY few, with their own implications (mostly performance wise). NAS is generally a MUCH better way also from security. You can not lock a client from looking at ALL data in a SAN if he can connect (he has block access), but a NAS can enforce file level security. – TomTom Jan 16 '13 at 12:06
  • 1
    I'd never use block based SAN - except some very dedicated situations like shared storage for systems especially tailored for it (mainly VMware or other cluster-aware FS) - or for a single-system access like adding another disk to a server via ISCSI or like. All other multi-system access should better be block-based. – Truebsalgeblaese Jan 16 '13 at 12:11
  • Yes. See my answer - it basically is also a security thing. Can you TRUST the client? If not - use a NAS (file level access). Even with a shared file system - a computer can just corrupt it on purpose or steal the block image to extract files later. – TomTom Jan 16 '13 at 12:12
  • 1
    Another answer that does not answer the question about WHY in depth. – TomTom Jan 16 '13 at 13:22
2

Now, we can also create file system on server operating system for block level storage, so what is the advantage of using file level access with NAS.

  • You are not limiting yourself to cluster aware file systems, which are few and with their own problems (the problem in iteself is complex).
  • A NAS has security on file level which is pretty important for any business level file access. In a SAN you can do that on the disc level, but every machine can read the whole disc, which means you can do a disc level analysis. A NAS can regulate access on the network interface efficiently.

That pretty much is it. Generally SAN is used for

  • Single computer access (remoet boot to a volume without discs internally)
  • Backend storage (no USER goes there, only computers)

A NAS on the other hand

  • Is used for variety user access as it has better security (can filter out files people are not allowed to see) in one file system.

Technically there is not a lot different. Many SAN can serve as NAS, pretty much every NAS can also work as SAN. There is no hardware that is required for a SAN - many people think SAN is Fibre Channel controllers. THat is "old tech". Yes, theey are dominant, but they are not particularly fast either. 10G, Infiniband are better (particularly infiniband) when you can use them. This really these days is more a logical separation.

THe most important thing in many cases is the trust boundary. Even with cluster file systems - it would be utterly stupid to hand out files via a SAN when every computer can see the WHOLE SHARED DISC and then download it and try to decrypt the file system. A NAS has a defined safe api here - you ask for a file, if the NAS says it is not there, you can not access the disc to bypass this.

TomTom
  • 50,857
  • 7
  • 52
  • 134
  • *Many SAN can serve as NAS, pretty much every NAS can also work as SAN.* - Ok. But when? Any examples would be great. – Drt Jan 16 '13 at 12:32
  • What you mean with "when". When you configure it. Have a NAS - Read the documentation. It 99% (unless old) can be used as as SAN via ISCSI. Same the other way. When? WHEN YOU SET IT UP. Use description - READ MY ANSWER. – TomTom Jan 16 '13 at 12:33
  • Sorry for not clearing. By 'when', I meant any real life examples of applications or scenarios which use this way? – Drt Jan 16 '13 at 12:49
  • As I said. NAS when you do not trust the client, SAN when you need shared block storage. THe problem with SAN is that to read from a SAN do you not only need a shared file system but also TRUST the client, as the client can see the whole disc. With a NAS, the NAS is the trust boundary. User has no rights to a file, he may not even KNOW it exists, and can not get it. – TomTom Jan 16 '13 at 12:54
  • Fibre Channel can be 16gb pushing up to 2GB/sec through each port. That's 60% more network speed than ethernet. FC is about speed, ethernet about convenience. Other than that I agree with you. – Nate Jan 16 '13 at 13:57
  • @Nate welcome to the modern world. Infiniband is now 40gigabit per port, for a cost lower than 10g ethernet. Latency blows you away - that is basically pci-4 (quad channel) cross rack. Put that in a SAN, put virtual machines around and you blow everything away for a VERY low cost. FC has it's point, but between 10G (price) and Infiniband (performance) it really is - hm - not always good. My las projet we pulled a constant 5GB/s over a infiniband link from storage. – TomTom Jan 16 '13 at 14:02
0

The primary advantage of NAS is that dissimilar devices can access the same filesystem at the same time.

David Schwartz
  • 31,215
  • 2
  • 53
  • 82
-1

Probably the most common file system in NAS is NFS, which can be accessed easily from several devices at the same time. There are solutions for multi-device-use also in SAN, but then you need global file system, which may be considered not that easy to set up.

grassroot
  • 683
  • 5
  • 14
  • 1
    -1. CHeaper? How? Most NAS devices this time also serve as SAN and are more than happy to dish out ISCSI access, which means that there is no differne. A SAN needing FC interfaces is hogwash - you can run that ofer Ethernet, 1G, 10G and for high speed you want something more powerfull anyway (like infiniband). NOTHING in the "SAN" word says you HAVE to have FC. – TomTom Jan 16 '13 at 12:07
  • Agreed. It is an "old school" way to use SAN with FC and it is not required nowadays. I edited my answer, although I think you explain the difference more comprehensively in yours. I did +1 that. – grassroot Jan 16 '13 at 12:56
  • Actually FC has TONS of disadvantages. It is nice on paper - but the F in it is particuarly hard to deal with. These fibers are tricky beasts - I take some copper over them whenever I can ;) Thanks. – TomTom Jan 16 '13 at 13:21