WAN optimisation is a well-established and popular technology.
Until recently, this was invariably done via dedicated hardware appliances and for connections between branch offices and main datacentres.
In this podcast, ComputerWeekly.com datacentre editor Archana Venkatraman talks to storage editor Antony Adshead about WAN optimisation, cloud optimisation and mobile optimisation and the key choices available to IT managers.
Venkatraman: What is WAN optimisation and why is it useful?
Adshead: WAN optimisation is aimed at organisations connecting remote locations and datacentres, and addresses inherent issues with WAN links, most notably latency and less-than-optimal bandwidth use.
These issues start with a simple shortage of bandwidth that costs money and often increases in inconvenient jumps, meaning the next upgrade requires quite a leap in costs. That means connections are often maxed out and organisations struggle to deliver good application performance due to latency, or fail to protect data adequately because the network is struggling to transfer backups.
Next, there is the problem that network protocols can be overly chatty, or that they don’t handle incomplete packet transfers well. Network traffic can end up looking like a sawtooth graph; traffic increases until it hits its max, then drops off significantly as packet retries occur until the graph rises again and maxes out again, and so on.
And there is also the fact that delivery of data is often duplicated; that is, more than one user in a location is likely to want the same data as their colleague, and there are better ways to deliver that than to keep sending it over the wire.
So, WAN optimisation products deal with these issues by a variety of means. These include local caching so that users in one location don’t repeatedly have to get data from elsewhere; optimisation of troublesome network protocols, by deduplicating repeated header patterns for example; data deduplication and compression to reduce the amount of data sent down the wire; and traffic shaping and prioritisation of data.
So, what’s it all good for? WAN optimisation allows you to send more network traffic more efficiently than you can over box-standard WAN links. It’s good for application delivery, for example, between a central datacentre and remote offices, and it’s also good for data protection and disaster recovery (DR); ie sending backup data back from remote locations to primary and DR datacentres.
Venkatraman: Who makes WAN optimisation products and what form do they come in?
Adshead: Dealing with the second part of the question first, historically the only way to buy WAN optimisation products was to buy hardware appliances. These come in various specifications of capacity and throughput and are installed at central and remote sites where they process network traffic moving between them.
Key suppliers include Riverbed, Cisco, Silver Peak, Blue Coat and F5. You’ll want to look at dedicated hardware appliances if your traffic needs are heavy.
Suppliers differ in their approach to WAN optimisation, utilising different techniques. So, for example, you may want to choose a product that emphasises local caching capability if you have sites that repeatedly use similar data sets. Or you may want the emphasis on TCP optimisation if your organisation depends on large amounts of random traffic.
The key disadvantage of dedicated hardware is, of course, that it adds more boxes that you’ll need to manage, and scaling up means adding hardware, but it may be unavoidable if you need lots of processing power applied to WAN optimisation.
More recently, another form factor choice has arisen – the virtual WAN optimisation appliance. These see the implementation of WAN optimisation processing capability via a virtual machine installed on a physical server.
These are available from most of the suppliers mentioned above, plus others. Their key advantage is that deployment is quick and easy, and scaling up by addition of further instances of the software is also relatively simple. All you really need to be certain of is that you have sufficient processing capability available on a physical host.
Herein also lies the key potential disadvantage. WAN optimisation depends on being able to iron out WAN inefficiencies by applying processing power to network traffic. But putting lots of virtual machines on one physical host will cause contention for processing power and throughput. Therefore, virtual WAN optimisation may well suffer in terms of efficiency, which kind of defeats the object of the exercise if it reaches certain levels.
So, likely scenarios where you would use virtual appliances are where you want quick deployment and just would not consider investing in hardware WAN optimisation.
Those are the key form factors in WAN optimisation, but there are others, as we’ll see.
Venkatraman: Where next; what about cloud WAN optimisation and mobile WAN optimisation?
Adshead: Two recent trends in IT more widely have pulled WAN optimisation along. These are the cloud and the proliferation of mobile and the bring-your-own-device (BYOD) phenomenon.
The rise of the cloud has brought the possibility of doing away with appliances – physical or virtual – as the location for WAN optimisation processing. Now, it’s possible for a third-party provider to optimise WAN traffic in the cloud.
Customers are provided with a VPN link into the cloud and the provider takes care of optimising traffic. Of course, you still need bandwidth to and from the provider, but the cloud service can take care of caching persistently accessed data at a point of presence close to your site. One example of cloud optimisation providers is Aryaka, which provides optimised internet connections via 25 points of presence worldwide.
Mobile optimisation, meanwhile, tackles the BYOD phenomenon, and is where a so-called “branch office of one” or an employee with a mobile device can have traffic to and from their hardware optimised via on-board software. Mobile optimisation products are available from the likes of Riverbed, as well as networking specialists such as Circadence.
Such products provide performance benefits, but are, however, limited in their capabilities compared with those suited to larger sites, as they don’t have multiple streams of data that provide optimisation opportunities, and of course they also contend for processing power on the employee device.