Perceived Performance - Virtualization Group - Boston

natureplaygroundΤεχνίτη Νοημοσύνη και Ρομποτική

14 Νοε 2013 (πριν από 3 χρόνια και 11 μήνες)

95 εμφανίσεις

Understanding

Tim Mangan

Kahuna
: TMurgent Technologies LLP

http://www.tmurgent.com

Microsoft MVP for Application
Virtualization

President
, Virtualization Boston
http://www.virtg.com


Computational VS Perceived Performance


What
D
o Users Care About?


Measuring Perceived Performance


Virtual Environment Performance




Measurement of discrete components of a system
that affect overall performance.

Examples include:


Processor time


Bandwidth


Queue lengths


Counters of all kinds

A methodology where one analyzes the system
with a goal of improving user productivity by
focusing on issues that affect the performance
as
perceived by the users.

Computational

Perceived


Repeatable


Break down to directly
measurable components



Variable


Measure “end to end”

Our systems today are too complex to rely on
computational performance (alone).


Multi
-
system


Networking


Remote Protocols


Virtualization Layering


Rethinking the Progress Bar
1

1 Harrison,
Amento
,
Kuznetsov
, & Bell
http://chrisharrison.net/projects/progressbars/ProgBarHarrison.pdf


Good:

Linear

Fast Power

Bad:

Any with variation




1

http
://
www.projectvrc.nl



Independent framework to measure virtual
machine performance


Free


Set up your own
tests


Understand their results


Virtual Session Index



Office 2007 SP1

SWAP Space (RAID
5
)

Add
XenApp

5.0

2003 versus 2008 x86

2003 versus 2008 x64


Bare Metal
vs

VMs

TS versus Desktops

Odd Results

Bare Metal
vs

VMs

Page Sharing

Virtual Desktops


Disk I/O Subsystem very important


Scenarios were CPU oriented


I/O subsystem of Hyper
-
V was bottleneck



New results being published soon


Improved Test Methodologies


New Vendor Releases


Nehalem Rocks for TS! 2x Improvement


Hyperthreading

great (except ESX)


EPT
-
D Rocks (Hardware
vs

Software)


i
f ESX and EPT
-
D, turn
vMMU

software off,


if ESX and no EPT
-
D, turn
vMMU

software on


New version Hypervisors did not improve much


Vsphere

4.0, Hyper
-
V 2.0,
XenServer

5.5


Hyper
-
V 2.0 improved more than others


Xen

& Hyper
-
V now almost identical, except that a single VM cannot
have 8CPU under Hyper
-
V (limit=4)


Vsphere

can be better/worse than others
depending on scenario


Office 2007 SP2 fixes “outlook preview pane” performance


IE8 performs on par with IE7

New!

Here are some links
most relevant to this
topic.

Project Virtual Reality Check:

http://www.projectvrc.n
l


Perceived Performance:

http://www.tmurgent.com/WhitePapers/PerceivedPerformance.pdf

http://
www.tmurgent.com/WhitePapers/PerceivedPerformance_VirtualOS.pdf


VDI versus TS (video):

http://www.brianmadden.com/blogs/videos/archive/2007/04/25/vdi
-
solutions
-
a
-
year
-
later
-
from
-
briforum
-
2007.aspx

http://www.brianmadden.com/blogs/videos/archive/2007/10/09/xen
-
versus
-
esx
-
a
-
performance
-
head
-
to
-
head
-
comparison
-
from
-
briforum
-
europe
-
2007.aspx

Citrix Logon and Logoff Chart:

http://
www.brianmadden.com/blogs/gabeknuth/archive/2008/08/14/briforum
-
video
-
the
-
excruciating
-
detail
-
of
-
the
-
xenapp
-
logon
-
process.aspx