From: www.itworld.com

Troubleshooting SaaS performance issues

by David Messina

September 23, 2008 —

Traditional network troubleshooting tools and processes were developed around the hub-and-spoke concept, with remote applications the spokes into servers in a centralized data center. But Saas applications that are delivered from an off-site provider’s data center defy traditional monitoring assumptions and require a new methodology for troubleshooting.

Challenges of troubleshooting SaaS performance issues
While network operations personnel have local visibility and control of applications deployed from within the enterprise, they are much more limited in SaaS applications served “from the cloud.” In particular, it’s difficult with traditional tools to distinguish SaaS traffic from other Internet usage going in and out of remote sites. IT personnel do not have local ability to determine whether the service is up and running or if the business user has straight connectivity to his or her hosted application. And while deep packet inspection (dpi) can deliver this level of granularity, this technology is costly to deploy across the distributed enterprise -- where most remote sites have direct Internet access.

Ultimately the challenge for an IT organization is that their end users aren’t going to call the SaaS provider if they have performance issues; they’re going to call the IT support desk.

Before support personnel relay the problem to the SaaS provider, they first need to perform the due diligence on their end to see if there’s anything in the enterprise affecting the SaaS application. For example, it could be a legitimate issue with Saleforce.com availability, for which IT personnel need verification to invoke the provider service-level agreement (SLA). Or the sluggishness could be the result of contention for the Internet.

This issue is compounded by the fact that SaaS users within the enterprise potentially could be working from disparate remote sites using different types of access and competing for resources, each with a different mix of applications, including peer-to-peer (P2P) applications that can rapidly saturate their link to the Internet. Or the end user might be on the wireless LAN and competing for shared bandwidth with other aggressive applications running that have fluctuating usage patterns, are experiencing slowdowns or have basic issues with signal interference.

New Visibility for networked applications
In order to troubleshoot SaaS performance issues, IT needs to understand the perspective of both the application and the end user and to see all the permutations inside and outside the infrastructure. That visibility is not available through traditional network management or monitoring applications. But new technologies are available that use information from network flows to discover all servers, networked applications, and user interactions with those elements by name and category.

Using this tracking information, profiles are created for all applications and for the networked application experiences of each end-user that enable IT to detect any unusual behavior and determine the root cause of the performance issue. In this way, IT can rapidly uncover the hidden causes of SaaS performance problems by determining which applications are generating high bandwidth on the link or by pinpointing unauthorized applications that are causing congestion.

With accurate information, IT personnel can quickly shut down problematic users under their control and alert the SaaS provider to any issues on their domain ... with the proof they need to get the problem resolved quickly.

David Messina is VP of Marketing at Xangati.