Brocade's File Area Network messages
Part 1: File storage network intelligence
I attended a Brocade File Area Networking (FAN) event and this is a report of what the audience of prospective and existing Brocade customers heard. (You can read about the origination of the FAN term here.)
Brocade sees itself as connecting, managing, and optimising shared storage IT environments. It is not a SAN-only product supplier. Paul Phillips, Brocade's UK and Ireland country manager, said: "We connect to many more server ports than storage ports. We want to build on that." It now has three lines of business: SilkWorm SAN products; Tapestry File Area Network products; and services. The FAN space is the growing one and Phillips said: "We're still out there looking for acquisitions in the file area space."
File data estate
Ivan Veall, FAN business development manager, said: "Typically there are separate storage devices for block and file data. We'd like to remove the block and file distinction in your file/data estate so you can just manage data. Personally I think data area networking (DAN) is where it's all heading."
"We need to apply structure and rigour to the file area network idea, the idea of a data centre fabric integrating file and block data management. A FAN is a collection of systems and software enabling centralised management of file data that drives user and administrator efficiencies. There are seven components:
1. End clients
2. Connectivity between clients and file namespace.
3. Policy-driven file management and control services
5. File serving devices
6. File systems
7. Storage devices.
Brocade has a collection of FAN software tools as a result of its NuView acquisition:
- MyView - software controlling access to files.
- UNCupdate - sets up links between applications and data that is being migrated.
- StorageX - a virtual file manager sitting on top of file data. It is OEM'd by Network Appliance as VFM.
- Tapestry WAFS - wide area file services based on Windows servers.
- FLM - file lifecycle management, an archiver.
- Data on Demand - a detailed classification tool.
Veall said: " We're adding a greater layer of intelligence over file servers. If you consider Microsoft's Distributed File System (DFS) then VFM is DFS on steroids."
Comment: This list, fitting in the 2-4 FAN components area above, ignores Unix file servers. Although Brocade didn't say anything about heterogeneous file area networks I think it would be curious indeed if Brocade did not have ambitions to bring Unix file servers and Unix client systems into its FAN world.
Next up on the Brocade speaker roster was George Mobargha, a solution architect. He said there is more file data than block data and it's harder to manage. Inactive file data consumes expensive disk resource. Think of the 80:20 ratio for inactive:active data ratio. File management today is carried out from a server-centric view. i.e. C:/ folder structure. This isn't a user view.
When data is moved form one disk to another then application links to those files are broken. Manual mapping is needed to add new capacity and accomplish load sharing, also disaster recovery is accomplished through manual procedures. These are the reasons why the FAN concept is gaining momentum.
The global namespace (GNS) is a key FAN concept. Think of it as a set of unique filenames that provides location transparency. Think of it as a kind of URL for files that is translated to a physical file address in the same way a Domain name Server translates a URL to an actual IP address.
An IP address is like a UNC path file reference in an application. IP is conceptually equivalent to CIFs or NFS. Users access a GNS name for a file which is mapped to a physical location for that file. The GNS presents a virtualised file system.
Comment: asked about single instance storage Mobargha said: "Yes, there is a need. We don't have it today however." He didn't know whether Brocade should build it, partner for it or leave it to the operating systems.
StorageX is our flagship product and manages centralised file storage. It is built on a global namespace. Brocade adds policy-based automation. "We can now consolidate file storage without affecting users. We can replicate from remote offices to the datacentre and thus provide DR for remote offices." Brocade can provide automated failover to a DR site when the primary site goes down. It can also do storage load-balancing. Brocade can supply data classification at a logical level.
Comment: all this competes with file-based storage vendor products such as ones from 3PAR, Acopia and Isilon, etc.
StorageX has a built-in replication engine. It can also leverage NetApp's SnapMirror technology,
UNCupdate is a software tool that helps in data migration by altering embedded file access links in a document or spreadsheet. It changes it from a physical reference to a logical reference.
FLM migrates file data from a primary NetApp device to secondary storage. It works only on NetApp today; it manages the file lifecycle on NetApp filers. Use it to identify and move inactive data from primary to cheaper secondary storage.
Inactive data classification is done by file extension; it's not content-aware; that introduces a CPU burden. Brocade's FLM can block file creation based on non-permitted file types using the file extension.
When files are moved to secondary storage stubs are left behind on the primary storage.
Comment: this is outside the global namespace capability. In that file mappings are altered when a file moves and there is no need for residual stubs. Accessing such a moved file requires extra disk accesses: Access primary and find stub; access FLM server and get moved file new location from stub reference; access new location on secondary storage. Leaving a stub is typically done by archiving software, not virtualised file software using a global name space. So Brocade's GNS doesn't embrace tiered storage.
No integration is planned to integrate FLM into the GNS capability, integrate FLM into StorageX, get rid of the stubs and extra disk accesses to locate a moved file. Mobargha says the extra disk accesses don't matter too much; it's aged and inactive data after all.
Part 2 of this article will continue describing Brocade's FAN products and strategy.