But is it the application itself, or other factors, that cause both a network (local and wide area) and individual workstations to perform below their potential?

Doug Scott, director of hosting services for MICROS-Fidelio, offers the following suggestions for users of the popular OPERA PMS running on a LAN. One of the biggest areas of concern to Scott is the quality of the internal cabling plant at the hotel.  Incorrect wire, improper termination and lack of patch panels are all items that can cause performance issues.  An improperly terminated data jack, while it appears to be working, can actually have loose connections that can cause re-trys that slow performance.  All terminations should conform to the TIA/EIA 568-B specification and use CAT5e- or CAT6-rated wire and jacks.

Switches and hubs should be of commercial quality and replaced as necessary.  Scott also emphasizes the need to keep spyware, malware and viruses off workstations.  These rouge programs may have gotten through the virus/spyware protection software, and, while not destructive, can degregate performance significantly.  Limiting access to the Internet by end users is one way to cut down on unwanted menaces.

Servers should be properly maintained and the disk space allocated to the databases, as well as overall disk space usage, should be monitored.
Saeed Karim, chief technology officer, hospitality solutions at SoftBrands emphasized both initial and on-going planning to maintain performance, both on-property and at central hosting sites. He suggested analyzing usage patterns, especially peak times, the mix of centralized vs. local applications, and the security requirements of each.

While the majority of applications are still locally based, more hotels companies are turning to centralized, Web-based data centers. The biggest area that causes poor performance of hosted applications is a lack of bandwidth at the property, or not enough bandwidth allocated to the central applications.  E-mail and Web usage can require significant resources so Web site access should be restricted, where possible.

Redundancy in connections is required in hosted applications and that redundancy should come from separate sources (not the same vendor, in case of a significant outage).  The size of the pipe into a data center must be adequate for now and in the future or performance can suffer.  All contracts negotiated with suppliers should include upward scalability if performance begins to decline.  This includes additional bandwidth or additional T1s or T3s.

Karim emphasized that if multiple servers are involved, whether on property or at a central Web server farm, they should be connected with a 1 gigabyte (GB) switch, instead of the more common 100 megabyte model.  This will increase performance significantly.

Scott offered this insight when it comes to the perception of system speed.  Hotel users that worked on the old character-based UNIX and DOS systems of the ‘80s and early ‘90s remember such systems as being fast.  There were no graphical interfaces, no connectivity to the Internet and no e-mail.  Most required the entry of codes several characters long (with no way to look up such codes).  While modern systems may appear slower than their predecessors, there is no comparison in functionality.

Geoff Griswold is a hardware and wiring specialist for the Omni Group. Geoff can be reached at theomnigroup@mindspring.com or (888) 960-8787.