Hardware Reference
In-Depth Information
6.22 [5/15/15/20/25] <6.5>
Realizing the Capability of Amazon Web Services
: Imagine you are the
site operation and infrastructure manager of an
Alexa.com
top site and are considering us-
ing Amazon Web Services (AWS). What factors do you need to consider in determining
whether to migrate to AWS, what services and instance types to use and how much cost
could you save? You can use Alexa and site traffic information (e.g., Wikipedia provides
page view stats) to estimate the amount of traffic received by a top site, or you can take
concrete examples from the Web, such as the following example from DrupalCon San
a-day.pdf
. The slides describe an Alexa #3400 site that receives 2.8 million page views per
day, using a single server. The server has two quad-core Xeon 2.5 GHz processors with 8
GB DRAM and three 15 K RPM SAS hard drives in a RAID1 configuration, and it costs
about $400 per month. The site uses caching heavily, and the CPU utilization ranges from
50% to 250% (roughly 0.5 to 2.5 cores busy).
a. [5] Looking at the available EC2 instances (
htp://aws.amazon.com/ec2/instance-types/
),
what instance types match or exceed the current server coniguration?
b. [15] Looking at the EC2 pricing information (
htp://aws.amazon.com/ec2/pricing/
)
, select
the most cost-efficient EC2 instances (combinations allowed) to host the site on AWS.
What's the monthly cost for EC2?
c. [15] Now add the costs for IP address and network traffic to the equation, and suppose
the site transfers 100 GB/day in and out on the Internet. What's the monthly cost for
the site now?
d. [20] AWS also offers the Micro Instance for free for 1 year to new customers and 15 GB
bandwidth each for traffic going in and out across AWS. Based on your estimation of
peak and average traffic from your department Web server, can you host it for free on
AWS?
e. [25] A much larger site,
Netlix.com
,
has also migrated their streaming and encoding
infrastructure to AWS. Based on their service characteristics, what AWS services could
be used by Netflix and for what purposes?
6.23 [Discussion/Discussion/20/20/Discussion] <6.4>
Figure 6.12
shows the impact of user
perceived response time on revenue, and motivates the need to achieve high-throughput
while maintaining low latency.
a. [Discussion] Taking Web search as an example, what are the possible ways of reducing
query latency?
b. [Discussion] What monitoring statistics can you collect to help understand where time
is spent? How do you plan to implement such a monitoring tool?
c. [20] Assuming that the number of disk accesses per query follows a normal distribu-
tion, with an average of 2 and standard deviation of 3, what kind of disk access latency
is needed to satisfy a latency SLA of 0.1 sec for 95% of the queries?
d. [20] In-memory caching can reduce the frequencies of long-latency events (e.g., access-
ing hard drives). Assuming a steady-state hit rate of 40%, hit latency of 0.05 sec, and
miss latency of 0.2 sec, does caching help meet a latency SLA of 0.1 sec for 95% of the
queries?
e. [Discussion] When can cached content become stale or even inconsistent? How often
can this happen? How can you detect and invalidate such content?
6.24 [15/15/20] <6.4> The efficiency of typical power supply units (PSUs) varies as the load
changes; for example, PSU eiciency can be about 80% at 40% load (e.g., output 40 wats
Search WWH ::
Custom Search