Tools, tricks, and lessons learned from those who support radiology, cardiology, and other imaging specialties throughout medical enterprises.
Gb networks and PACS workstation performance.
As existing networks are being expanded and sometimes replaced, it is common for gigabit network (Gb network) to be used as a replacement alternative to connect PACS workstation (its availability in server rooms has been common for many years) over older 100 megabit networks. A gigabit network corresponds to a theoretical data throughput of 1000 megabit per second (not 1000 megabytes per second as it is sometimes mistaken for), which is a 10 fold increase over the next lower network performance level, 100 megabit per second.
The potential benefit of a gigabit networks for PACS workstation has been increasing over the years, as the amount of power of the various components of PC’s has been steadily increasing, so that the speed at which they process data is much greater than before, and workstations end up waiting after the network to do their job. At the same time, the data explosion mostly brought about by multislice CT has provided ample justification for the increased workstation capability, thus leaving the older network speed as being the weakest component, not having kept pace with just about everything else. As a rule of thumb to best appreciate the situation, a 100 megabit network speed corresponds very roughly to a throughput of slightly more than 10 megabyte data per second.
Given that a CT slice is 0.5 Mb, this gives a transfer rate of 20 slices per second, this being optimistic as the theoretical maximum speed is typically not achieved for a variety of reasons. Knowing that the size of modern CT studies can be up to thousands of slices, it is obvious that a 100 megabit network speed poses a considerable challenge for the handling of such large studies. This can be sometimes mitigated by different predictive strategies to somehow anticipate the need to bring certain studies beforehand, but it is unavoidable that even in the best of cases, that need will not always be expected (especially if radiologists want to read from many possible locations), so that the result of the above transfer time calculation will be often inflicted upon users.
Going from 100 to 1000 megabit (1 gigabit) per second radically changes the situation. In practice, the apparent 10 fold increase may not be fully realized since other bottlenecks will then kick in and start playing a bigger role. For instance, the workstation hardware will be much more challenged at 1 gigabit, despite the larger capability mentioned previously, but it is likely that the transfer rate will be manyfold nonetheless, thus resulting in a dramatically improved user experience. Therefore, bringing network gigabit capability to workstations is sure to allow users, especially radiologists, to be put in a much better position to handle large CT studies, although the associated storage problem will obviously remain.
The implementation of gigabit networks of course requires an appropriate network infrastructures (the “switches” to which the network connection are wired), but this may not be the main problem in terms of cost. The cables running from the workstation network hookups to these switches also need to be of particular type. For workstations, these cables are almost always of the type “Unshielded twisted pair”, or UTP, made of 8 copper cables bundled into a single containing wire. Such cables themselves come in various categories, referred to as CAT followed by an alphanumerical code. Better categories are characterized by more protection against external radiofrequency noise, which make them more suited for larger network speed.
Gigabit networks require at least CAT5e cabling, while legacy cabling may be of lesser type, such as CAT5. Even the above mentioned CAT5e may prove insufficient when the lengths involved between the hookup and the switch are large, so CAT6 cables, an even better type, may be required. Whenever cabling is found to be inadequate for the deployment of gigabit network, new cables will need to be deployed, and it is the labor costs associated with these rather than the cost of the cables themselves that will constitute the bulk of the project cost. The associated work can also be quite invasive, and impact activities, unless performed outside operation hours, thus bringing about an even greater labor cost.
Despite these inconveniences, the benefit calculated previously will undoubtedly provide enough justification for bringing gigabit speed to PACS workstations, and any new planned facility, or even physical any renovation project where PACS workstations are present, should incorporate the deployment of gigabit networks to connect these workstations.
Alain Gauvin, McGill University Health Center