HP dc73 HP Blade Workstation Solution Planning Guide - Page 25

Roundtrip, Latency, Subjective impact, Behaviors, Network hops, Geographic distance

Get HP dc73 - Blade Workstation Client PDF manuals and user guides View all HP dc73 manuals
Add to My Manuals
Save this manual to your list of manuals

Page 25 highlights

Based on actual measurements at a number of blade workstation installations, RGS shows a maximum frame rate between 150 FPS and 200 FPS, suggesting that the RGS processing time becomes non-negligible in lower- latency network environments. There are also practical limits to interactivity. Given that the typical LCD monitor has a refresh rate of 60 Hz, it is not necessary to drive the RGS frame rate beyond this. Doing so would be expensive, and would not benefit the user. There are several other points to consider: • Human visual acuity starts to perceive smooth motion, rather than distinct steps, between 10 and 15 frames per second. • Most streaming video is displayed at either 15 or 30 frames per second. Table 5-2 details typically observed latency values and their effects Table 5-2 Latency values and their effects Roundtrip Latency Subjective impact < 20 ms (local LAN) Just like local 20-80 ms (regional) Interactive (user may notice slower update rates) 80-200 ms End user will know system (transcontinental) is remote (occasional use) > 200 ms (intercontinental) Need to evaluate use case Behaviors • 15-30 fps (MPEG video encoded at 20 fps; full motion video at 30 fps) • Frame rate update typically driven by native application performance • Interactive models may be choppy (5-15 fps) • May be slight drag in window drags • Occasional clicks and pops in streaming audio • Acceptable for applications that do not drive high frame rates (

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
Network Planning 25
Based on actual measurements at a number of blade workstation installations, RGS shows a maximum frame rate
between 150 FPS and 200 FPS, suggesting that the RGS processing time becomes non-negligible in lower- latency
network environments.
There are also practical limits to interactivity.
Given that the typical LCD monitor has a refresh rate of 60 Hz, it is
not necessary to drive the RGS frame rate beyond this.
Doing so would be expensive, and would not benefit the
user. There are several other points to consider:
•
Human visual acuity starts to perceive smooth motion, rather than distinct steps, between 10 and 15 frames
per second.
•
Most streaming video is displayed at either 15 or 30 frames per second.
Table 5-2 details typically observed latency values and their effects
Table 5-2
Latency values and their effects
Roundtrip
Latency
Subjective impact
Behaviors
< 20 ms
(local LAN)
Just like local
•
15-30 fps (MPEG video encoded at 20 fps; full motion
video at 30 fps)
•
Frame rate update typically driven by native application
performance
20-80 ms
(regional)
Interactive (user may notice
slower update rates)
•
Interactive models may be choppy (5-15 fps)
•
May be slight drag in window drags
•
Occasional clicks and pops in streaming audio
80-200 ms
(transcontinental)
End user will know system
is remote (occasional use)
•
Acceptable for applications that do not drive high frame
rates (<5 fps)
•
Mouse interactions can get out of sync, leading to
situations like over-rotation of CAD objects.
•
Streaming audio may lag video
> 200 ms
(intercontinental)
Need to evaluate use case
•
Recommend piloting for specific applications
To maximize RGS interactivity, the goal should be to minimize network latency. The key factors that determine
network latency are:
•
Network “hops”—
The number of devices (such as routers and switches) between the RGS Sender and RGS
Receiver should be minimized. Each time the data goes through a switch, latency is added to the
connection. The latency added by each switch is dependent on the type of switch or router used.
•
Geographic distance
—Every 100 miles (160 km) between the RGS Sender and Receiver adds
approximately 1 millisecond of roundtrip latency. Furthermore, as the distance between the Sender and
Receiver grows, there are typically more switch and routing devices in the network path. While it is not
always possible (nor desirable) to keep the blade workstation and client close together, intelligent analysis of
routing tables can prevent a cross-town connection from first crossing the continent.
•
Network saturation
—Heavily loaded networks can dramatically increase latency. Buffering and dropped
packets cause not only increased latency but also cause inconsistent latency (jitter). The resulting interactivity
will be uneven, which may be more objectionable than sluggish but predictable behavior.
•
Quality of Service (QoS) Rules
—In addition to latency added by physical devices, latency can be affected by
Quality of Service (QoS) rules and packet-size queue priorities defined for the routers that connect the RGS
sender and RGS Receiver. It is not unusual for smaller packets to be routed to high-priority queues, while
larger packets are given lower priority. Packets can also be tagged with QoS information to enable them to
be routed with varying priority.