• Post Reply
  • Bookmark Topic Watch Topic
  • New Topic

Scalability VS Reliability

 
s khosa
Ranch Hand
Posts: 72
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Hi,

Horizontal Scaling is supposed to increase relaibility. It sounds fine to me. But why would Vertical Scaling reduce reliability??? I read on a book about this statement. Is it because we consider aggregation of CPU failure rates when we add multiple CPU's to a box? Any pointers!!!

Thanks
 
Required Field
Ranch Hand
Posts: 39
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
I wonder why. The way I understand it, Vertical scaling should not reduce reliability. It doesn't increase reliability but nor should it reduce reliability. I think it should stay put (constant).

Theoretically ofcourse. I am not the best reliability guy so don't go by my word alone!



Originally posted by s khosa:
Hi,

Horizontal Scaling is supposed to increase relaibility. It sounds fine to me. But why would Vertical Scaling reduce reliability??? I read on a book about this statement. Is it because we consider aggregation of CPU failure rates when we add multiple CPU's to a box? Any pointers!!!

Thanks
 
Peer Reynders
Bartender
Posts: 2968
6
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
In essence:

Horizontal Scaling (parallel redundancy) - A single point of failure takes out one machine - the others can still continue to work. Note however that you also need to take precautions in the front-end, persistent-store and interconnections so that a single point of failure can't disable the entire system.

Vertical Scaling - Usually makes an existing system more complex, adding more potential single points of failure that are capable of disabling the entire system.
[ February 21, 2006: Message edited by: Peer Reynders ]
 
Sreenivasa Majji
Ranch Hand
Posts: 224
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
When you encounter this kind of question, think what effects the given concept (such as Reliability). As Peer mentioned, single point of failure effects reliability. Adding more RAM or CPUs makes system bit more complex doesn't decrease chances single point of failure. This is doesn't mean that it reduces the reliability, it should stay as it is.

Hope this kind of ambigious question doesn't appear in actual test.
 
Raghubir Bose
Greenhorn
Posts: 8
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
I would rather put it this way.
Reliability is defined as "The probability of failure-free software operation for a specified period of time in a specified environment".

With vertical scaling, meaning with the addition of CPU, the performance increases , but the probability of failure does not change(neither increases or decreases).
 
Peer Reynders
Bartender
Posts: 2968
6
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Originally posted by Raghubir Bose:
With vertical scaling, meaning with the addition of CPU, the performance increases , but the probability of failure does not change(neither increases or decreases).


Vertical Scaling is a very general term.
Addition of CPUs is the most popular form of vertical scaling. However there are other ways of adding hardware to increase performance.

Adding a CPU into an empty slot of a well designed multi-processor server should not decrease reliability. In may increase performance depending on the operational profile.

Moving from a single CPU system to a multi-CPU capable system with only one CPU equipped could reduce reliability as there are more controllers that could fail on the multi-CPU system.

Adding a hard drive to a single hard drive system for interlaced operation can improve performance and can be seen as vertical scaling for disk bound applications, however it will reduce reliability as a failure of either drive will be disasterous. You will have to add two more drives mirroring the first two drives to increase reliability; however your reliability may still be below the original single drive system because of your dependence on more complex (fail-over) controllers.
 
  • Post Reply
  • Bookmark Topic Watch Topic
  • New Topic