LostKiwi
Senior Member
- Joined
- Aug 25, 2006
- Messages
- 31,343
- Reaction score
- 21,601
- Location
- Midlands / Charente-Maritime
- Your Mercedes
- '93 500SL-32, '01 W210 Estate E240 (RIP), 02 R230 SL500, 04 Smart Roadster Coupe, 11 R350CDi
One of the big issues with the cloud is latency, i.e. the time required for a single request to transit from host to cloud and back. This isn't normally an issue for single request transactions but when an application makes 90 requests as part of a transaction it becomes a bigger issue.
For example if a request takes 20ms to make the return trip from end to end and back (at that's quite low for the interest) a large transaction comprised of 90 requests could take 1800ms which could be a major issue. It's not quite that simple (buffers, TCP sliding windows etc all mitigate this to some extent).
For example if a request takes 20ms to make the return trip from end to end and back (at that's quite low for the interest) a large transaction comprised of 90 requests could take 1800ms which could be a major issue. It's not quite that simple (buffers, TCP sliding windows etc all mitigate this to some extent).