Database Reference
In-Depth Information
Table 4.3
Average error (rates) of high stress database read over different time periods
Cloud
Database
Server
Connection error
Successful
databases
error (%)
error (%)
error (%)
request (%)
Amazon simpleDB
0:00 (0:000)
0:00 (0:000)
41:00 (0:127)
32;359:00
(99:873)
Amazon localDB
0:00 (0:000)
16:40 (0:051)
6368:40 (19:656)
26;015:20
(80:294)
Microsoft windows azure
table storage
0:00 (0:000)
0:00 (0:000)
11;593:80 (35:783)
20;806:20
(64:217)
Google datastore
2:25 (0:007)
4:75 (0:015)
5462:75 (16:860)
26;930:25
(83:118)
Table 4.4
Average error (rates) of high stress database write over different time periods
Cloud
Database
Server
Connection
Successful
databases
error (%)
error (%)
error (%)
request (%)
Amazon simpleDB
111:17 (0:343)
9:50 (0:029)
2470:83
(7:626)
29;808:50
(92:002)
Amazon localDB
0:00 (0:000)
25:20 (0:075)
5262:60
(16:243)
27;112:20
(83:680)
Microsoft windows azure
table storage
0:00 (0:000)
0:17 (0:001)
4810:33
(14:847)
27;589:50
(85:153)
Google datastore
31:67 (0:098)
3037:37 (9:374)
4787:50
(14:776)
24;543:66
(75:752)
smaller percentage trend in reading than writing, while Microsoft Windows Azure
Table Storage and the local database in Amazon EC2 on the contrary, display higher
rates in read operations than write operations.
Amazon SimpleDB via Amazon EC2 maintains the lowest error rates in both
reading and writing, almost approaching 0% in read tests. While the local database
via Amazon EC2, which shares the same instance with the web application of
Amazon SimpleDB via Amazon EC2, started receiving a high percentage of
connection errors from 1,500 concurrent requests. The reason of this phenomenon
could be explained by that the local database causes additional resource contention
by virtually being inside the same instance as the host server instance. This leads
to a less scalable architecture, as a trade-off to smaller latency from host server to
cloud database.
For Microsoft Windows Azure, the connection error percentage begins to leap,
from less than 1% at 1,500 requests, to more than 50% and 30% in reading and
writing separately at 3,300 concurrent requests. This indicates that a limit in terms
of what this Azure server instance can handle has been hit.
For Google App Engine, a large number of connection errors under high load
has been observed. Most connection errors from Google App Engine contain the
access denied message, which is a standard HTTP 401 error message. Through
cross checking the server side, there is no record of HTTP 401 at all in the Google
App Engine. This means that these requests are blocked before getting into the
