WAN optimisation can be used to ensure than network performance is not impacted when storage infrastructures are decentralized and data sits at a main data centre and at satellite offices. Another alternative would be to upgrade the network, but this is a potentially expensive proposition given the big steps in which bandwidth tends to jump from 10 Mb to 100 Mb, and from 1 GB to 10 Gb.
WAN optimisation tools work by reducing the amount of bandwidth required to send data over a wide-area network between users at separate locations. Some products compress the data so it takes up less space before staging it to local cache files. Vendors offering this approach include Cisco Systems, Expand Networks and Juniper Networks; these products work best with data-centric tasks such as remote file access or data replication.
The adoption of WAN optimisation is now moving into the mainstream, with 40% of enterprises implementing them in production mode and another 2% testing them, said Whiteley.
WAN optimisation deployments tend to take place in three phases. The first phase often involves optimising network performance between a head office and remote sites experiencing the most problems in terms of application response times. The second step usually sees the technology deployed between the primary data centre and a secondary disaster recovery (DR) site to enhance data backup and recovery speeds, while the third phase entails rolling out WAN optimisation tools across the organisation's remaining sites.
"From a storage point of view, if organisations are having data protection issues with their remote sites, this is a good option to consider as it introduces control and enables people to introduce better policies that are consistently and centrally maintained," said Hamish Macarthur, chief executive at analyst firm Macarthur Stroud International.
These benefits are possible because Space Engineering Services now has a better understanding of the type and volume of traffic between its two main offices in Bristol and its 11 remote sites. Using a Web front end to the Expand Networks' Compass offering that it deployed at the start of 2008, Space Engineering Services can view when the network links at each facility are underutilised, hitting capacity or somewhere in between.
The tool works by compressing data and sending it over the network to a cache disk that stores it. This means such technology tends to be best suited to groups of users that access the same applications-based data repeatedly, and it works particularly well on point-to-point leased lines. It is less effective with email traffic, as messages change each time and are therefore less suitable for caching.
Adam Bess, IT manager at Space Engineering Services, explains the benefits. "Because of the granularity with which we can see traffic moving up and down the network, we saw in one instance that links were getting maxed out as someone kept trying to access a large financial spreadsheet and it was taking a lot of bandwidth," Bess said. "So we re-assessed where the file should best be stored and that rectified the problem."
A key issue is that most of the firm's office-based staff and data are located in Bristol, which means heavy traffic to and from these sites due to the centralised nature of many of the company's databases, files and other systems such as email. Space Engineering Services is currently employing the information provided by its WAN optimisation tool to undertake a review to help it understand "where data is and should be held at each end of the network" Bess said.
Such information will be useful as the business opens more new offices during the year. "We'll be able to spec and deliver the most appropriate architecture for that office based on the work that's done there," Bess said, "as we'll have a better understanding of what the implications will be."
Another benefit of WAN optimisation technology is that it has enabled Space Engineering Services to set quality-of-service rules on traffic such as file transfers. If key personnel need to access critical documents, they can now be given priority over less-important activity.
But, Bess warned, "it's important to get users from across the business to agree to priorities in order to gain a consensus." For example, he said, if bandwidth should max out when the rules are implemented, "people will know that access to certain files or services will be slower than others and won't be in a position to complain."
Space Engineering Services embarked on its WAN optimisation project to improve application response times, which had slowed due to increased volumes of traffic on the network resulting from a combination of business growth and increased Internet usage.
In attempting to address the issue, Bess was keen to avoid the significant expense of having to upgrade all of the firm's BT MPLS network links. When one of the company's partners, Esteem Systems, recommended Expand's WAN optimisation offering as a way to speed up and provide better visibility and control of traffic passing over the network, Space Engineering Services decided to undertake a trial.
"It was very useful to prove that the concept worked and to demonstrate the benefits to users," Bess said. "This helped justify the cost and make the business case to some offices. It also gave us a better idea of the applications that were being used most to determine whether compression and acceleration would be enough to avoid upgrading the network."
Once given the go-ahead, the offering took only 20 days to roll out as "the hardware just sits on the network, so it's a case of sending it to each branch and getting it plugged in," Bess explained. "Introducing WAN optimisation technology has enabled us to focus [our] expenditures on where we had issues and where there was a clear business case. So we hit our objectives of better visibility and increased speed, rather than just throwing money at the network."
The firm regularly stores huge amounts of data, which includes drawings of up to 50 Gb in size. These files are accessed regularly if a contract is won, and must legally be retained for the lifetime of the structure; if the tender is unsuccessful, however, the files can be archived.
Given the critical nature of this data, it had become clear that both failover and DR provisioning needed to be in place to ensure that such information was retrievable at all times. The challenge was how to do it cost effectively.
Seamus Conway, network manager at Rogers Stirk Harbour & Partners, says one of the options they looked at was whether to use a leased line to connect the company's headquarters to its DR provider's site. However, the idea was abandoned because the necessary 100 Mb link would be too expensive.
"We started evaluating alternatives and when we put in the WAN optimisation boxes from Riverbed, we found they squeezed traffic down by 90%, which means that we've been able to use a standard 2 Mb line," Conway said. "It's much cheaper, which is an important consideration, as disaster recovery services aren't cheap. But it also means that even in the first year, we're seeing a payback."
Rogers Stirk Harbour & Partners opted for the Riverbed Technology product because the company believed a traffic shaping-based offering was more suitable to its requirements than a caching-based one. According to Conway, the latter choice is "like sticking a chunk of disk online and giving people access to the data. But the problem is that, if you lose it for some reason, you have to repopulate the entire disk with a new copy of that data."
Conway said a traffic shaping-based system is preferable because it "improves network speed and latency in a way that other products don't, so if you want to increase throughput, it works very well. Also, in a worst-case scenario, if the box blows up, the data patterns can be recouped quickly and data synchronisation is very quick."
An important consideration in this context is to undertake a sizing exercise to understand how the organisation works with its data, how many people require access to it and what device size is required.
"A sizing exercise dictates which model to go for," said Conway. "It's easy to make the wrong choice. None of these solutions are cheap, so you don't want to get it wrong and end up buying something small, then have to upgrade and find you wasted money with the first purchase because you can't use it any more."
Other things to consider relate to product failover options. Such options enable remote sites to access their own locally stored patterns of data should the connection with the main site go down. Rogers Stirk Harbour & Partners decided not to use such functionality because it would mean that two different groups of people could end up working on two different versions of a file, which would not be satisfactory.
Users need to think very carefully about failover features during procurement, Conway said. "Because much of the disk is segmented for failover use, if you chose not to use it, you can't reclaim that portion of the disk for online data storage."
Rogers Stirk Harbour & Partners is considering whether to centralise its IT and storage infrastructure to its London headquarters before deploying WAN optimisation boxes at its Tokyo, New York, Madrid and Hong Kong offices to ensure staff can access data as quickly as if it were sited locally.
This was first published in February 2009