Peer-to-Peer Storage of Electronic
R. Rudin, P. Szolovits & A. Advani
In 2004, President Bush initiated a movement to establish a nation-wide
system of interoperable electronic medical records (EMRs) by 2014. How
will these records be stored? Current EMR storage and back-up technology
does not adequately account for disasters. Recent disasters such as Hurricane
Katrina resulted in delays in restoring medical data, ranging from days
to weeks. Katrina has disproved the rule of thumb that it is sufficient
to have only one back-up site located 25 kilometer from the primary site.
We aim to demonstrate that peer-to-peer storage based on a Distributed
Hash Table (DHT) implementation (also developed at MIT's CSAIL) is a suitable
method for storing and backing-up EMRs. We demonstrate this by simulating
the storage load of a typical hospital network and measure the following:
- Latency for reads and writes
- Data loss in the case of disaster
We also compare these results with a master-slave implementation and
show that peer-to-peer storage on DHTs has characteristics befitting EMRs.
We argue that this method, if optimized, would be better than the master-slave
implementations currently used.