Vous êtes sur la page 1sur 2

Assignment 4

The 700,000 hour MTBF = (# of devices tested * # of hours tested) / # of failures.
700,000 operating hours / 2000 disks = 350
If 2000 disks are used for 350 hours that gives you 700,000 hours of operation, and with a 700,000 MTBF
1 disk failing is expected.
You can expect a disk to fail every 350 hours.
Operating hours * failure rate = # of failures, so reducing operating hours reduces the number of failed
disks. 2000 disks and 700,000 hours MTBF assuming the disks are under the same usage and conditions
as when the MTBF was calculated. High humidity, dust, and other factors can cause the disks to fail
faster, thereby increasing the failure rate. All you can do to reduce failure rate is have the disks running
in optimal conditions.
Disks fail so to ensure zero data loss, data redundancy is needed between the disks, so if one fails then
that same data is on another disk so nothing is lost.
Virtualization is running an operating system on software, and using it as you would a physical computer.
When looking to increase workload, virtualization is used because managing virtual machines is much
easier than dealing with additional physical counterparts. Also, virtual machines are isolated, so if one
machine is infected than the other machines are unharmed. Cloud computing uses virtualization to shift
workload from the users computer to other machines.
Multiprocessor systems split workloads between multiple processors and clusters split workloads
between multiple tightly coupled systems. In a cluster, the multiple systems used together have the
same hardware and operating system, whereas in a grid the computers are loosely coupled and can
have different hardware and operating systems in combining computing power.
Multiprocessor systems execute parallelizable code much quicker than a single processor when a
majority of the code is parallelizable. A disadvantage is that there will not be the speed up you expected
after adding processors if a small portion of the problem is parallelizable and the majority of it is
sequential. Cluster computing involves tightly coupled computers so there is little latency and better
performance compared to loosely coupled grid computing. An advantage of loosely coupled systems is
that any system can be added to the grid regardless of operating system or hardware, so creating a grid
is cheaper than a cluster which requires specific hardware and operating system requirements.

To ensure data security, data must be redundant across multiple disks so if one disk fails no data will be
lost because that same data is stored on another disk. Backups on multiple sites are needed in case an
entire site is compromised, i.e. a fire occurs in the data system, so no data is lost even in a severe case
such as this. In order to implement this system, RAID disk arrays can be used since it allows disks to be
hot-swapped, ensuring minimal downtime and provides data redundancy across an array of disks. Any
RAID level other than RAID 0 can be used depending on the specific needs of the data center. RAID 0 is
not advised although it is the best performing of the RAID types; it does not provide protection against
disk failure because there is no data redundancy.