Taming Your Flock of NAS Devices By Ray Lucchesi published: Friday, October 05 2007
NAS devices are easily deployed but capacity limited, leading to an administratively unmanageable number of NAS devices as mount/share points multiply. This administrative quagmire is further complicated with a multi-vendor NAS data center where cross-vendor functionality is often lacking.
File virtualization services (FVS)
Thus enters a new class of products, collectively termed file virtualization services, to mitigate the problems of proliferating NAS devices. Many vendors have entered this market with products such as F5 Networks Acopia ARX Intelligent File Virtualization, Brocade StorageX Global Name Space and EMC Rainfinity Global File Virtualization.
Most of these new products solve the problems of NAS device proliferation and excess redundant capacity with critical functionality. Perhaps most importantly, the new FVS products support advanced name space functionality. Less advanced products use shared name spaces, which introduce an additional mount/share point. In contrast, the more progressive products use federated name spaces, which offer a joined name space across the NAS devices to be virtualized but does not introduce a new mount/share point. Equally important, the new FVS products are further differentiated by operating either in-band, out-of-band, or a combination of both to provide their advanced capabilities.
Additionally the new FVS products maximize usage of existing NAS device facilities by providing transparent (to the user and the application) file directory migration. One helpful by-product of transparent directory migration is storage tiering, where less active data can be moved off high performance NAS devices to lower performing NAS devices.
The new FVS products also support data replication across heterogeneous NAS devices making a multi-vendor environment more efficient and disaster recovery more effective. For advanced data centers, file virtualization products also enhance performance load balancing in which active directories are moved from busy to idle NAS devices in real-time. File performance and throughput can be measurably improved.
As expected, vendors have taken a disparate path to provide their own unique solution. Below we briefly describe three of the top-selling products available on the market today and how they compare under differing operational scenarios.
FVS product descriptions
EMC offers Rainfinity Global File Virtualization. This appliance sits on the network and operates in-band during advanced file virtualization services and out-of-band once these services are completed. For example, during directory migration and file name remapping the device is in-band but it returns to out-of-band when these operations are completed. Thus, Rainfinity only minimally impacts performance while supporting full native NAS box functionality. Rainfinity uses a federated file name space. This product offers a clustering option with a pair of servers for high availability.
Brocade offers StorageX Global Name Space products. This software product sits out-of¬band and operates as a Windows directory server. It uses Windows Distributed File System (DFS) services to offer advanced name space services to Windows clients. Because it is exclusively an out-of-band solution, file directory migration for open files uses snapshots followed by re-synchronizing all changes to the migrated data when files stop being accessed. Policy support for migration, storage tiering, and remote replication somewhat eases this constraint by automating these processes and retrying file synchronization for open files until they are closed. StorageX also uses a federated file name space. This product offers a clustered servers option for high availability support.
F5 Networks offers the Acopia ARX Intelligent File Virtualization product line. This appliance sits in between clients and NAS devices introducing a new mount/share point which operates in-band at all times. While in-band operations can support all file virtualization functionality, it does introduce another mount/share point and may negatively impact NAS performance, limiting native NAS device functionality. F5 Acopia ARX provides a shared file name space. ARX supports clustered configurations for high availability. Additionally, due to its in-band operations, different performance levels are offered to support varying data center NAS performance requirements.
OPERATIONAL SCENARIOS
Normal NAS operations
Operating as separate mount/share points, in-band FVS products such as F5’s Acopia ARX must intercept all NFS and CIFS requests and re-route them to the correct NAS device. Furthermore, products using this architecture introduce another directory layer into the network so they inherently must manage directory information at both the NAS device level and the shared name space level. This adds overhead and complexity to all file create and delete operations. To minimize performance impacts, F5 ARX uses a dual-path architecture which routes file data requests down a fast-path and control requests, needing more processing, down a slower path through the same appliance.
Brocade’s StorageX product is only accessed when the client’s DFS referral cache is out-of-date. However, because this product is out-of-band, migration re-synching must wait until files are inactive or closed. Any space or performance advantages to be gained from the file migration must await completion of this resynchronization process.
EMC’s Rainfinity product attempts to provide an optimum solution to minimize performance impacts while still providing transparent active directory migration. The Rainfinity product does this by selectively operating in-band during advanced functionality, returning to an out-of-band state when the activity is complete. Thus, NAS performance is only minimally affected while active files are transparently migrated and the directories remapped. However, Rainfinity’s solution adds code complexity by changing from out-of-band to in-band and back again to reduce this performance impact but this is completely automatic and invisible to the user. Rainfinity neither adds overhead or complexity for file creation or deletion.
Virtualizing NAS devices
In-band FVS products like F5’s Acopia ARX require prolonged startup activity to populate their shared name space with file location information when incorporating active NAS devices. They mask this activity by first incorporating the files being accessed and assimilating inactive files in the background, but this may still impact initial file access until its directory entry has been processed. Brocade’s StorageX product essentially supplies DFS directory services and also requires prolonged activity to virtualize active NAS devices. EMC’s Rainfinity product, due to its hybrid in- and out-of-band approach bypasses this directory assimilation process and only requires a one-time network configuration to virtualize active NAS devices and is non-disruptive to end-user data access.
NAS advanced features
As F5 Acopia ARX creates a new-shared name space layer in the network it renders inoperable some add-on NAS device advanced features. For example, FlexVol a capability of NetApp NAS devices that allows file systems to dynamically expand, becomes unusable. Native NAS device snapshot is also unusable. However, F5 Acopia ARX provides their own snapshot services, which may someday support cross vendor snapshots, but cannot take advantage of the space efficient native NAS device snapshots. EMC and Brocade do not need to support cross-vendor snapshots because they fully support native NAS device/OS advanced features such as NetApp FlexVol and snapshots.
NDMP, a capability that supports NAS device outboard backup/restore services, can be made unreliable by F5 Acopia ARX. By introducing a shared name space layer, the physical connection between files and the directory path in which they appear can be broken. This causes NDMP to miss migrated files during backup of NAS devices. In contrast, for federated name products like EMC and Brocade file data always resides in the identified directory path. Hence, with EMC, and Brocade, NDMP can be used to back up migrated data for NAS devices as easily as data that has never been moved.
Appliance outages
The inherent architecture and reliance on shared name space mode of in-band FVS products such as F5’s Acopia ARX increases the risk of a complete NAS outage if the appliance becomes inoperable. They have introduced a shared name space layer, that when crippled, disconnects all clients from the file services provided by the underlying NAS devices. Data centers have a lengthy, multi-step process to re-establish access to physical NAS devices. Original NAS device mount/share points must be re-established then all NAS devices must be scanned to locate migrated data and their new mount/share points activated. Returning to F5 Acopia ARX or shared mode virtualization is yet another multi-step process.
In contrast, Rainfinity loses all advanced functionality when inoperable, meaning no further file migration, storage tiering, and remote replication can occur, yet there is no single-point of failure that could bring down NAS devices or client access to the file system. File migration operations that were in progress are aborted and the files are left intact at the source location. However, while Rainfinity is down all normal file level access is retained, all current mount/share points are still active and the end-user can still perform all file operations to access and update NAS files.
StorageX, when inoperable, also loses advanced functionality. By using DFS referral caching, all current file access is available and unaffected, however future file access may be impaired when the DFS referral cache is out-of-date.
Summary
FVS products have emerged to help data centers administer the rapidly increasing number of NAS devices as well as enable cross-vendor NAS management. Each of the products addressed above by F5, Brocade and EMC offer much of the same benefits, but differ significantly in their approach to providing these benefits.
Each product has advantages and disadvantages over its competitors. Making the ultimate choice requires careful consideration of the anticipated:
-
Overall NAS performance required
-
Number of NAS devices being virtualized
-
Usage of advanced NAS device features
-
Concern for prolonged file access outages.
About the author
Ray Lucchesi is president of Silverton Consulting in Colorado. He has worked in data storage for over 25 years mostly in marketing, engineering, and directing new enterprise class disk and tape automation products.
At Silverton, Ray has helped both Fortune 500 and startup storage vendors improve product development and marketing. He also helps data centers select and use data storage more effectively. Ray speaks at major storage conferences and writes articles for storage publications.
At StorageTek, Ray focused on marketing and product strategy as well as ILM product architecture. Before that, he directed embedded code development for StorageTek’s enterprise disk subsystem. Much earlier, Ray was lead architect
and later managed the team that developed the host software for StorageTek’s first automated tape library.
Ray holds a BS from Bradley, is an officer of the IEEE Denver Section ExComm and is also a registered member of Technical Advisory Service for Attorneys.
www.silvertonconsulting.com
|
