Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
Taming branch office sprawl has become something of an obsession for both enterprises and the vendors who serve them, with much of the emphasis on technologies such as wide area file services (WAFS), WAN optimization and application acceleration that can provide centralized control of data.
But are those technologies the best approach for everyone?
Arun Taneja, consulting analyst and founder of the Taneja Group, says there are a number of issues to consider in managing remote branch offices: loss of control of data, security, application performance, network bandwidth costs, duplicate equipment, personnel costs and regulatory compliance.
According to Taneja, the choices are: centralized storage with all data served over the network to remote users via WAFS, WAN optimization, compression and other similar technologies; or a distributed architecture, with arrays, tape drives and other storage gear used at all sites.
As for which is best, there is no easy answer. One thing is for sure, however: Remote offices are here to stay. Taneja estimates there are already two million in the U.S., containing 75 percent of all employees and a whopping 60 percent of created data.
"You must either improve the management of a distributed architecture, or consolidate and serve everyone from a central location," says Taneja.
Let's take a closer look at both options.
Centralize and Optimize
The argument in favor of centralization is straightforward. When each branch and remote office runs its own storage and servers, it becomes difficult and costly to manage, protect, backup and maintain. Further, competent IT personnel are required at remote sites. It's costly to pay for them at every location, yet it can be disastrous to omit them when servers and arrays have to be maintained. Stories abound of backup goofs committed at branch offices by the uninitiated.
"A centralized architecture means all the data and services are where they can be more easily administered, a great benefit when there is no tech support on site," says Mike Karp, senior analyst at Enterprise Management Associates.
One of the touted benefits of centralization is lowered IT costs. Instead of having duplicate hardware at all locations, one data center houses the expensive gear, with distant sites maintaining a much lighter profile.
While centralization sounds good, though, it is fraught with challenges.
"The benefits of centralization are dependent upon remote users being able to gain access to centralized data and files at near-LAN speeds," says Mark Urban, director of marketing at Packeteer.
One of the big problems is overly chatty protocols such as TCP, CIFS, MAPI, HTTP and FTP. Taneja says this can mean 100 round trips for a simple action such as a file open. This translates into 10 seconds to open a file regardless of the bandwidth.
A multitude of tools have entered the market to address this and other issues related to centralized storage. According to IDC, the WAN optimization market alone is now over $300 million, and Packateer is the dominant vendor. IDC predicts the market size will double by 2009.
"Our iShared products enable branch office storage and server consolidation, improved application performance, and centralized management for distributed enterprises of all sizes," says Urban.
The iShared product line consists of iShared Appliances for the main office, iShared Flexinstal software installed on existing server equipment, and iShared Mobiliti software to consolidate and control laptop and desktop data at the edge of the network. This, says Urban, offers the remote workforce data access without sacrificing data security or file consistency.
Packeteer WAFS is built on a distributed file system and utilizes branch office caching, latency and bandwidth optimizations to deliver rapid file access over the WAN. In addition, ARC (Acceleration, Reduction and Compression) WAN Optimization accelerates TCP over the WAN along with stream-based compression. The technology incorporates print, domain controller and networking and management services that enable companies to deploy fewer servers at lower cost in the branch. The appliances start at $4,600, and iShared Mobiliti software is $157 per user.
Packeteer, then, has evolved its early WAFS offering into one with much broader scope, a trend other vendors are following. IDC says the trend is for WAFS tools to be combined with other products that compress data streams, monitor traffic flows, prioritize traffic via QoS policies, and manage applications from a protocol perspective.
"The WAN optimization market is morphing into wide-area performance," says IDC analyst Vincent Lui.
In addition to Packeteer, other proponents of centralized tools include Cisco Systems, Expand Networks, Orbital Data, Riverbed Technology and Brocade.
"These firms have differing emphases, so evaluate closely," recommends Taneja.
Like its centralized counterpart, the advantages of distributed storage can be laid out in a convincing fashion. By having the data spread around multiple sites, superior disaster recovery capabilities are available. By eliminating dependency on data communication lines to transport all data to a central location, faster data access is available at the remote sites.
"Consolidation can create problems such as the core applications becoming too slow over the WAN," says Taneja. "Remote users can sometimes feel like second-class citizens to the main office."
He remains in the centralized camp, however. His view is that WAFS and its partner technologies address most of the issues that plague a centralized architecture. He concedes that a distributed approach could be made to work, but that the cost is probably too high.
The cost side, however, can be mitigated by technologies such as Storage over IP (SoIP).
"SoIP is a cheap way to distribute storage widely," says Karp.
One of the principal advocates of this approach is Zetera Corp. While IP SANs require dedicated hardware, SoIP doesn't. It addresses groups of disk blocks as IP addresses. By directly mapping IP addresses to each storage element, this eliminates the need for dedicated controllers or aggregators. This also adds virtualization capabilities without the presence of additional virtualization software. Users can stripe, mirror and transfer data without regard to the location of the physical drives.
"Our Z-SAN allows storage to be added to any location, and then each site can access storage at the other locations," says Jeff Greenberg, senior director of product marketing at Zetera. "The reliance on an IP network means that the system can be installed and run on the organization's existing infrastructure, eliminating the need for specialized complex and expensive equipment like Fibre Channel switches and TCP offload engine NICs."
Z-SAN technology consists of two major components: a client driver that interprets file system commands and converts them to block-level Z-SAN network transmissions; and a network adapter that receives Z-SAN commands and converts them directly into disk-oriented I/O commands/payloads.
But this is far from the only method of bolstering branch office performance under the distributed model. A wealth of disk-to-disk technologies, says Taneja, aid in this regard. Companies such as Availl, Avamar, Data Domain, ExaGrid and Unitrends offer such products, he says. Note, too, that many of these tools can also be harnessed to improve the performance of a centralized architecture. They can either be used to make it cheaper and simpler to operate the distributed model, or faster and easier to serve files from a central location.
Data Domain, for instance, offers the DD400 enterprise series. The basic idea is to not waste storage space by holding the same e-mail attachment or document copy multiple times. Instead, pointers are used to replace redundant copies so that only one or two need to be stored. Further, the backup data stream is adjusted so that maximum data flows in the smallest possible unit of time.
The software built into the DD400 views data at a sub-block level. It looks at the data stream, and if that data has already been seen, it isn't backed up again. It is a disk-based backup system that also makes it easier to replicate the data over a WAN to a disaster recovery site. The company claims replication can be achieved with about one-tenth the bandwidth of before.
Remote Office Success
While opinions differ as to which approach is best, analysts seem to agree on a couple of points: that managing remote offices is no longer optional, and that no one technology will solve all the problems.
Taneja offers a couple of tips in the selection process.
"Identify your business issue in detail, rather than your technology choices," says Taneja. "Once you have accomplished that, take an 'arsenal' approach to combining the right technologies."
For more storage features, visit Enterprise Storage Forum Special Reports