Application service provision seems to offer a winning proposition. You get to use applications without upfront purchasing or licensing costs, and without the hassle of server maintenance or upgrades. All you need is relatively dumb terminals that can access a wide area network (Wan) connection. For a fixed fee per user each month, the application service provider (ASP) does the rest.
However, as you close the door on one set of problems, another is hurled through the window wrapped around a brick. Quality of service (QoS), reliability and security become key concerns, as well as how delivery over a Wan will affect the performance and functionality of applications. Application service provision is a relatively new way to buy software, and vendors are striving to resolve such problems.
So are the benefits to be gained by renting applications outweighed by the potential technical problems posed by this emerging technology, or would it be wise to wait until ASPs are well established?
Raising the ASP standard
Only one year old, the ASP Industry Consortium (Aspic), has 400 members including Microsoft, Lucent and PSINet and aims to tackle the technical conundrums of application provision.
Among its priority initiatives is producing a guide to service-level agreements (SLAs), being worked on by its Best Practices Committee. The group defines an SLA as a legally binding contract between the ASP and customer that specifies the agreed service quality. But as Aspic has found, defining and gauging quality of service is complex.
The consortium defines three main elements - network, system and application services - that should be encompassed in a single end-to-end SLA. This includes factors such as network availability, acceptable packet loss, latency service delivery timetables and management, and support functions such as helpdesk availability and a named account manager at the ASP with defined responsibilities to the customer.
Aspic quoted benchmarks from Stanford University's Linear Accelerator Centre's work on high-speed internet links. The consortium rates packet losses between 2.5 and five per cent as poor, and concluded that delays should be less than 250ms. Availability of Wan links should be to the highest telco standard.
Planet Online, a division of UK telco Energis that will launch ASP services in August, quoted 99.98 per cent uptime, while UK ASP Esoft Global, which is already up and running, quoted 97 per cent.
How or whether these guidelines should be incorporated into SLAs is a contentious point. Paul Webb, director of service operations at Esoft, one of Aspic's founder members, said that SLAs should spell out commitments in terms that network managers will understand.
"We don't see any reason to give users an SLA that they can't measure and understand," he says. "Packet loss, in our sense, affects screen response times. Outside of that, it's not an important factor of the service. I wouldn't expect my technical account managers to turn up to an account meeting and be beaten around the head because the packet loss was 10 rather than five per cent."
Spelling out expectations
What Esoft's SLAs do include are elements such as the percentage-availability of an application; agreed timescale for adding users; maximum allowable downtime; and the speed of response and availability of helpdesk services. More technical factors such as latency and packet loss are policed internally by the ASP, says Webb.
While an SLA should ideally follow the model outlined by Aspic, Todd Krautkremer, vice president of marketing for bandwidth management company Packeteer, says it's impossible, even for just the network element.
"The first phase of the market has been that a couple of ASPs have made feel-good claims of end-to-end QoS guarantees and SLAs," he says. "The reality is that you can't control what you don't own. If you are guaranteeing the performance of traffic over an infrastructure you don't own, then that simply is not an enforceable SLA."
Service provider Equant, which is selling commercial services including ASP on a global network originally built by the world's airlines, is a case in point. While it owns much of the network between its data centres, where applications are stored, and its customers, the service occasionally has to traverse the public internet.
"We'll be pretty upfront," says director of marketing Ian Lee-Emery. "We'll say we manage the SLA up to the end-point of where we can control, but we won't offer an SLA across the internet. We're happy to make hosting centres or networks available for a certain percentage of the time, but we couldn't possibly take responsibility for doing the same with the internet."
For larger customers, where service delivery travels over its own managed network and into the customer's site via a leased line, Equant does offer end-to-end SLAs. This is where the network operators that have moved into ASP provision believe they have stolen a march from other providers that have to strike deals with separate network service providers.
Andy Irvine, marketing and sales director of Planet Online, says: "One of the key things we've got is the Energis network and Planet Online working together. That's what gives us the advantage and allows us to control service level. That will be critical for end users."
ASPs that don't own their own networks may not be prepared to offer such watertight guarantees. Much depends on the relationship with their network service provider. Maynard is also founder of NetStore, which has been providing ASP services since March 1997 and now boasts 560 customer organisations, including Cisco and cosmetics company Avon.
The company discourages dial-up access, preferring to deliver via leased lines. Its ISP is NetComm, although there is no formal partnership between the two. NetStore does offer end-to-end SLAs, Maynard says, but the customer's own choice of ISP has to be taken into account.
"We would look at which ISP the customer is using and see if it has one-to-one peering with our supplier. If they do, then we say that's fine. If it's someone that doesn't, we might have to downgrade some of the service issues because we don't believe they've got the best connectivity," says Maynard.
The worst case scenario for companies that want to rent applications is when each part of the ASP service is owned by a different supplier. "Many ASPs that I've come across don't even take ownership of the application itself. It's a number of hand-offs - if its an application problem, a network problem or a server problem, they're all separate helpdesk phone numbers," says Lee-Emery.
Even in the best-case examples, where the ASP takes responsibility for delivering over a managed network and a leased line right into the customer's data centre, there could be problems, says Krautkremer. "In the whole ASP model, what's missing is this notion of a demarcation point: an application-aware point that defines the boundary of the ASP service offering. It defines where the ASP's responsibility ends and the customer's infrastructure begins."
Packeteer demonstrated a product called AppVantage at the Networks Telecom show, which it said can do just that, and which is being trialled by US ISP FutureLink. Sitting on the customer's premises at the point of entry of the ASP circuit, which might also carry the customer's ISP traffic, it offers per-user analysis of ASP sessions in terms of response time, latency, availability and so forth. It also distinguishes between such aspects of the ASP and the customer's networks.
"The customer will be able to get SLA compliance reports and application performance reports that show them exactly how their applications are performing and whether or not the ASP is meeting its SLAs," says Krautkremer. The best ASPs will offer this as part of their service, he believes, making their SLAs workable and enforceable.
The issue of the impact that ASP services has on customer networks was taken up by Mike Lucas, technology manager of Compuware, whose EcoScope product monitors and analyses network traffic at the application level and generates performance reports.
This is the sort of tool that ASP customers will need to properly assess how Lan performance is affected by ASP provision, says Lucas. It might be possible to save 25 to 40 per cent by renting rather than buying or licensing an application, but it will still consume Lan bandwidth and have to co-exist with existing on-site applications.
"You might think, 'if I rent an application, I don't have to make any investments'. But you will have to investigate and understand the impact of that ASP application on your Lan," says Lucas. "It might be that you are not be able to use the application because it will perform so poorly."
EcoScope might be offered by ASPs as an SLA watchdog, or customers might want it to police the service, adds Lucas. Either way, such functionality could be used to help ensure the success of the arrangement. "Both parties are going to have to spend a lot of time predicting how links will be delivering in terms of response time, capacity and latency, otherwise the whole thing will fold like a house of cards."
However, Tim Pickard, director of market development at Esoft, argues that the impact of ASP delivery on the Lan will be nothing to worry about. "It's relatively minimal compared with local delivery. If the customer has a 10Mbps Lan, they wouldn't have to think of moving to 100Mbps. If all the applications were being delivered locally, that may well be the case."
Time waits for no Wan
The speed of the Wan link that connects end users to the ASP server can also be a critical factor affecting performance levels. The modification of mainstream applications to fit the ASP model has to take this into account: the terminal-server connection is no longer a 10Mbps Lan segment, but typically 256Kbps or 512Kbps over a private circuit.
A common method of adapting applications, such as Office or Outlook, to ASP delivery is to wrap them up in Citrix or Microsoft's terminal emulation middleware Terminal Server. Citrix demands a per-user channel of 25Kbps over Wan links and is sensitive to latencies that exceed 90ms, according to Esoft technical design director Mark Nethercott.
That suggests that such leased-line capacities - and even the ISDN dialled back-up offered by NetStore - are adequate. But it's a tight squeeze and could restrict service complexity, so most ASPs are looking forward to the availability of digital subscriber line access over BT lines, which is expected next year. This promises speeds of several Mbps at less cost than the equivalent private circuit, and will offer a significant price/performance boost to ASP service.
Pickard believes the impact will be huge. "It's got all the things an ASP needs: fixed, low-cost access, and bandwidth is not going to be an issue."
Although ASP offers some immediate benefits - chiefly lower software costs for smaller firms - it still awaits software that best fits its architecture, as well as faster access technologies, to achieve its true potential. In turn, users must be sure exactly what the ASP is committed to deliver and may face extra complications in traffic management as a result.
As Krautkremer puts it: "In this new world of ASPs, the accountability does not change. An IT executive will get fired just as fast if an ASP fails as if they themselves failed."
BT wants to make the public switched telephone network history within eight years
Personal data being purloined by third parties via Facebook Login API
MacOS and iOS are better off apart, says CEO Tim Cook
Or they'll no longer be entitled to updates and bug patches