If you are a large enterprise customer with many sites or with many data centres sending large amounts of data over the WAN, you need to learn about WAFS. WAFS, or Wide Area Files Services, is a relatively new technology for an old concept. The goal of WAFS is to make file services and links across the wide area more efficient.
Networking professionals are not normally involved in decisions regarding file servers and data caching, but this case is different. WAFS devices sit at the edge of WAN links and, at some point, will require networking professionals to be involved in either implementation or troubleshooting. This alone should put WAFS on the networking radar.
Different vendors have different approaches for WAFS, but the goals are the same — to move data across links from 10x to 100x (or more) more efficiently than it is moving today. Many new possibilities for services and applications are created with this kind of WAN improvement.
So how does this work? In the past, data compression was used to help WAN links carry more data. This only saves a limited amount of bandwidth. WAFS uses data compression with other techniques to improve efficiency. The most common techniques are stream/data compression, data caching, protocol optimization and application optimization.
Today’s data compression techniques are similar to the past — redundant patterns of data are recognized and replaced with shorter tags as place holders for the full stream. On the other end, you recognize the tags and re-insert the full stream. So if you see a stream of 100 “0”s — you send something meaning 100 x “0”.
The next technique is data caching. Network data moving between sites, in many cases, is repetitive. A document is opened, basic changes are made and then the document is saved. Most of the document is identical except for the changes. Enter WAFS data caching. The first time the document is sent across the link, the WAFS system analyzes it for patterns that can be stored. The WAFS cache creates a tag for this data and stores it on both sides of the connection. The next time the same data is to be sent across, the tag is sent instead. The data can be in the same file or some other file (e.g. the same document attached to an e-mail). At the remote end the tag is recognized and replaced with the real data. This technique saves significant amounts of bandwidth when the same data is sent back and forth.
Protocol optimization works at the network and transport layers of the OSI model by recognizing chatty protocols that can have their overhead reduced. The WAFS system learns what communication must take place and what can be spoofed by looking at different parts of a network conversation. Spoofing is the process of doing local acknowledgements in a protocol, rather than sending the request all the way to the end device for an acknowledgement. There are a number of opportunities in today’s protocols to make improvements. The WAFS vendors have made their systems pretty smart and indicate that there is no tuning required. But, if you are going to tinker with communications protocols, you should understand what is going on under the hood.
Application optimization is the same idea as protocol optimization. It takes specific applications that endlessly communicate back and forth, and reduce their traffic on the link through spoofing and other techniques. By only sending smaller tags representing the communications, you free up bandwidth. This is especially true in applications such as NFS or CIFS, which allow you to browse network drives and send essentially the same data repeatedly. Again, you may not need to tune anything here, but you should understand what is going on.
This is a new market, but the vendors are lining up to get in. The two big networking vendors, Cisco and Juniper have each bought startups. Actona and FineGround went to Cisco, while Peribit went to Juniper. However, they are not the only game in town. There are any number of other companies in this field, with impressive performance numbers and features. Some of the more notable companies are F5, Riverbed, Tacit (recently bought by Packeteer), Expand Networks in partnership with DiskSites, and Availl.
The question is not whether this technology works, but whether it is useful for you. The vendors have all shown significant improvements in performance in model networks. Network traffic profiles, while similar from company to company, are relatively unique. They are based on how your business and IT organization has structured the data and where it resides. With centralized data, many branches and high latency between sites, WAFS may be a great solution. In the case of decentralized data, small transactions between sites, and no strain on the WAN, spend your money elsewhere. The only way to tell for sure is with a pilot. Make sure to try different vendors as well, since each vendor uses slightly different techniques, in different combinations.
Device failure is a concern. If a device fails on a link, the link should continue to function. Failures can be hard, such as a power supply failure, or soft, such as an OS or software bug or crash.
While hard failures do happen, the soft failures are more typical. Make sure there is a good answer for both. While this is not an exciting technology area, it is definitely a useful one. WAFS is a great technology that solves some very real problems for customers that fit the WAFS profile.
You should evaluate different vendor offerings to determine the actual amount of benefit, and make sure that your links are protected in case of failure.
If your company is having issues across the WAN and is looking at more bandwidth, this technology is definitely worth a look.