In-Depth
Acopia Networks Brings Threads Together
The mirror cracked ... thanks to suspicious information
The word “acopia” is derived from the Spanish expression “to come together;” it also has meaning in psychiatry: “the inability of a patient to cope.” Acopia Networks was experiencing an “inability to cope” with my earlier column discussing file system virtualization alternatives. The company asked to meet to correct misapprehensions that may have been left by the column.
At first, their request seemed less than compelling: most vendors want to rebut statements made about them in the media by a competitor. Rainfinity’s Jack Norris did have some unkind things to say about Acopia in his quoted remarks.
However, while a Rainfinity quote was included in the piece, we did not necessarily see the need to obtain and print a return volley from Acopia Networks. That type of “he said, she said” is the sort of thing you read in other publications.
Besides, Norris’ comments did not detract, in our view, from the main point of the piece, which was that there were many roads to file system virtualization. Moreover, the appearance of so many competing products in the market was making it increasingly difficult to differentiate one from another. What provoked writing about this subject in the first place was an inquiry from a reader, who included an e-mail dialog that showed the convoluted thinking process involved in puzzling through the alternatives.
We agreed to a meeting and discovered that Acopia’s boggle was not (only) about Norris’ comments, but errors of fact contained in the message thread we published. It turns out that the traffic forwarded to us by our kind reader, which contained a debate over the efficacy of an Acopia Networks solution versus a Network Appliance Virtual File Manager (VFM, a re-branding of NuView’s StorageX product), was not an internal e-mail thread at all. With the best of intentions and for our edification, the reader had lifted the dialog from a message board site for users of Network Appliance gear (toasters.mathworks.com), but in his zeal, had failed to notify us of this.
The message thread provided an example of consumers struggling with file system virtualization alternatives. Acopia pointed out to us, however, that one of the voices in the dialog was not a user but a vendor— someone who, at the time, worked for NuView. Some of his comments and observations were, in Acopia’s view, stilted, if not deliberately deceptive—especially those concerning Acopia Networks.
Under the circumstances, we agreed to give Acopia Networks equal time. After an hour-long discussion, we were glad we did.
First things first. Contrary to descriptions in the earlier piece, Acopia Networks does not offer “appliances.” The company makes “network switches,” according to Brendon Howe, VP of marketing and business development. “Only our entry-level product leverages an underlying commodity server chassis and might be properly regarded as an appliance. Our competitors make software that runs on a commodity Dell platform; we make a switch.”
Howe said that Acopia’s technology can be best described as a “file-level router.” It is installed between servers and storage repositories (NFS and CIFS file servers and NAS appliances) where it interprets I/O between clients and servers and routes file requests and responses to their appropriate destination.
Clarifying a point on which just about everyone was confused, Howe stated that Acopia switches do not impose a file system on files stored on NFS and CIFS storage devices. Rather, the product “pulls together the file systems of various file storage repositories and builds an aggregated namespace from them.” This namespace remains constant as changes are made to the physical devices on the backend. In other words, you can add capacity, move capacity around, or make other changes to the backend infrastructure while providing a consistent presentation of file folders to the end user or application.
Acopia’s recent announcement, which started the whole debate, highlighted two distinguishing characteristics, according to Howe: (1) its ability to create a global file system namespace across multiple copies of its switches that supports failover, and (2) its ability to load balance storage resources based on performance and latency characteristics.
The first is important because failover will not require the remounting of NFS (and arguably CIFS) volumes and the downtime that such dismounting and remounting entails.
The second feature is key, according to Howe, in cases where high-performance applications need optimal quality of service from their storage targets. Load balancing, Howe admits, is not a feature used by a majority of Acopia users. He notes that those with the most demanding applications need load balancing to address capacity or performance constraints. The majority of customers “don’t put enough load on our equipment to tax its throughput,” but when they do, the load-balancing capability is there, he says.
The mainstream use of the product is in data migration and back-end management. With its ability to provide a persistent presentation of back-end storage as storage volumes are added or removed, as data is migrated, or as systems are decommissioned, the product is the storage administrator’s best friend. Plus, Acopia has added tools to help automate such activities, which takes the drudgery out of capacity management and further helps to eliminate downtime.
Howe told me that many companies are currently using Acopia's products in conjunction with upgrading their Network Appliance Filers to 7G, requiring disruptive data migrations. Important is the product’s ability to eliminate the disruption of these migrations, and to provide the ability to redefine volume and “qtree” layout. (“Qtrees,” or Quota Trees, are a mechanism for partitioning traditional volumes into chunks of space with different security types and may be used to implement quotas on users and to implement other security policies.)
Howe refrained from making comparisons between Acopia’s switches and competing products. In its mainstream use, he noted, the most frequent competitor for Acopia are the professional services organizations of storage equipment producers, which sell file system management and tuning as a service.
Acopia’s switches come in three flavors: $25K buys a low-end unit, the ARX500, which can handle “a 100-million file infrastructure with approximately a gigabit per second throughput.” The middle-range product targets an infrastructure of about four times the number of files with roughly 2.8 Gbps of throughput, while the high-end platform scales up to a billion files and delivers up to 10 Gbps of throughput.
Howe says his speeds and feeds are the best in the business. We followed up on the issue raised in the discussion thread: how difficult is it for a customer to get a product to test? In the message thread printed previously, comparisons were made between the ability to download software from the Web and to try it before you buy it versus the need to pre-qualify for a demonstrator switch from Acopia and then wait in line for the next available box for testing.
Howe responded that it is, of course, much easier to download software than to request demonstration gear. “We have over 150 [demonstration] switches installed around the world right now that are chasing purchase orders.” To get on the list for a tester, he says, you need to have a “funded project.”
Asked why Acopia hasn’t gone the route of a software solution that can be mounted on commodity server hardware (as many of their competitors have), he stated that it was an early design choice. However, he alluded to future plans that might see a software-only version of Acopia wares.
So, in this column, we have done what we believe to be the correct thing: giving equal time to a vendor who viewed itself as unfairly slighted in a previous column. The real moral of the story is this—we welcome reader feedback on this column, but if you are providing information gleaned from other sources rather than your own personal experience, please let us know.
Your comments as always are appreciated. [email protected]
About the Author
Jon William Toigo is chairman of The Data Management Institute, the CEO of data management consulting and research firm Toigo Partners International, as well as a contributing editor to Enterprise Systems and its Storage Strategies columnist. Mr. Toigo is the author of 14 books, including Disaster Recovery Planning, 3rd Edition, and The Holy Grail of Network Storage Management, both from Prentice Hall.