Page 169 - How Cloud Computing Is Transforming Business and Why You Cant Afford to Be Left Behind
P. 169

D ANGERS ABOUND: SECURITY IN THE C L O UD



                 Internet in the northern Virginia area. The outage started at
                 12:34 a.m. Pacific time, according to Apparent, 34 minutes
                 before the dashboard reported a problem. It ended 44 min-
                 utes later, at 1:19 a.m. Pacific. The posts to the Service Health
                 Dashboard indicated an ongoing problem until 1:51 a.m.,
                 when the post, “The underlying power issue has been ad-

                 dressed. Instances have begun to recover,” appeared. The ini-
                 tial notice was late, but to be fair, the notice of recovery trailed
                 the start of the actual recovery by 32 minutes as well, accord-
                 ing to Apparent’s information.
                     The information that Amazon makes available when such
                 incidents occur seems to be aimed at minimizing the problem
                 rather than acknowledging its scope. It may be put there by

                 someone who is busy solving the problem, not by someone
                 who is on standby with nothing to do but explain mishaps to
                 the public. In short, Amazon Web Services achieves a high de-
                 gree of translucency with its Service Health Dashboard, but full
                 transparency is too much to expect from your cloud provider.
                 This is very different from data center operations, where the
                 people who answer questions can be fired by the people who
                 are asking them. In the data center, obviously, the facilities are
                 directly under a company’s control.

                     On the other hand, Amazon Web Services representatives
                 may debate my statement that postings tend to minimize the
                 severity of the problem. The second notice, at 1:26 a.m. Pacific
                 time, stated that the data center managers were experiencing
                 “power issues for a subset of instances in a single availability
                 zone.” This sounds like a contained outage, possibly a minor





                                                                     149
   164   165   166   167   168   169   170   171   172   173   174