HCloud, a Healthcare-Oriented Cloud System with Improved Efficiency in Biomedical Data Processing - Cloud Computing with e-Science Applications - page 188

TABLE 8.6
Network Traffic and Performance Statistics for Activities

Bandwidth
(KB/seconds) Receive/Sent Activity Total (MB)

size_rcv (size of responses) Uploading 835.84 224.58

Browsing 26,277.12 6,113.28

size_sent (size of requests) Uploading 2,999.04 803.37

Browsing 522.24 108.54

Source: Redrawn from He, C., Fan, X., and Li, Y., IEEE Transactions on

Biomedical Engineering 60, no. 1 (January 2013): 230-234.

TABLE 8.7
Statistics for Reading Performance of Amazon S3
Page Size (KB) 10 100 1,000

Response time (seconds) 0.14 0.45 3.87

Bandwidth (KB/seconds) 71.4 222.2 258.4

Source: Brantner, M., Florescu, D., Graf, D., Kossmann, D., and Kraska, T.,

in Proceedings of the 2008 ACM SIGMOD International Conference
on Management of Data , Vancouver, B.C., Canada, June 9-12, 2008,
251-263.

Time and Count rows. A response time less than 1 second is tolerable for the
user's experience. The actual number of concurrent users was nearly 30,000
during the experiment, as the Finish_users_count column of Table 8.3 and
Table 8.4 shows. So, it is acceptable to actual applications.
By comparison, Table 8.7 shows the reading performance of Amazon's
S3 with different page sizes (writing data to S3 takes about three times as
long as reading data) [27] . Ignoring the slight differences of hardware, the
usability of our system is satisfied in terms of Bandwidth , which is at least
double that in S3 (compare the last row of Table 8.6 with that of Table 8.7).
Response Time is also better, obviously. Overall, it demonstrates that this
document-oriented cloud storage architecture is more appropriate for the
health care services environment with a large number of trivial files rather
than non-document-oriented ones such as S3. Furthermore, due to platform
capabilities of linear extendibility, we simply increase the number of servers
if there is the potential for more users' requests in the future.

8.4.3 Comparison of Computing Overhead

From Section 8.4.2, we can find that the response time is acceptable to our
applications from the point of view of accessing HTTP. Actually, the main
tasks of HCloud are physiological data processing and computing, as dis-
cussed previously, which can affect the performance of the whole system.
We also take the ECG raw data process as an example. Using a typical parallel