diff --git a/windows/application-management/app-v/revision-heidi/appv-capacity-planning.md b/windows/application-management/app-v/revision-heidi/appv-capacity-planning.md index c4aa1ccf09..d416b882fc 100644 --- a/windows/application-management/app-v/revision-heidi/appv-capacity-planning.md +++ b/windows/application-management/app-v/revision-heidi/appv-capacity-planning.md @@ -79,349 +79,23 @@ The following table describes each factor that impacts round-trip time in more d |The number of connection groups configured on the management server.|For up to 100 connection groups, there is no significant change in the round-trip response time on the publishing server. For 100–400 connection groups, there is a minor linear increase in the round-trip response time.| |The number of access groups configured on the management server.|For up to 40 access groups, there is a linear (approximately 3×) increase in the round-trip response time on the publishing server.| - ---- - - - - - - - - - - - - - - - - - - - - -
Factors impacting round-trip response timeDescription

The number of publishing servers simultaneously requesting package metadata refreshes.

-
    -

  • A single management server can respond to up to 320 publishing servers simultaneously requesting publishing metadata.

    • -

  • Round-trip response time for 320 pub servers is about 40 seconds.

    • -

  • For <50 publishing servers simultaneously requesting metadata, the round-trip response time is <5 seconds.

    • -

  • From 50 to 320 publishing servers, the response time increases linearly (approximately 2×).

    • -

The number of connection groups configured on the management server.

-

-
    -

  • For up to 100 connection groups, there is no significant change in the round-trip response time on the publishing server.

    • -

  • For 100–400 connection groups, there is a minor linear increase in the round-trip response time.

    • -

The number of access groups configured on the management server.

-

-
    -

  • For up to 40 access groups, there is a linear (approximately 3×) increase in the round-trip response time on the publishing server.

    • -
- The following table displays sample values for each of the previous factors. In each variation, 120 packages are refreshed from the App-V management server. |Scenario|Variation|Number of connection groups|Number of access groups|Number of publishing servers|Network connection type|Round-trip response time (seconds)|Management server CPU utilization| |---|---|---|---|---|---|---|---| -|Publishing servers contact management server for publishing metadata at same time|0
0
0
0
0
0|1
1
1
1
1
1|50
100
200
300
315
320|LAN|5
10
19
32
30
37|17
17
17
15
17
15| -|Publishing metadata contains connection groups|10
20
100
150
300
400|1
1
1
1
1
1|100
100
100
100
100
100|LAN|10
11
11
16
22
25|17
19
22
19
20
20| -|Publishing metadata contains access groups|0
0
0
0|1
10
20
40|100
100
100
100|LAN|10
43
153
535|17
26
24
24| - - ---------- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
ScenarioVariationNumber of connection groupsNumber of access groupsNumber of publishing serversNetwork connection type publishing server/management serverRound trip response time on the publishing server (in seconds)CPU utilization on management server

Publishing servers simultaneously contacting management server for publishing metadata.

Number of publishing servers

-
    -

  • 0

    • -

  • 0

    • -

  • 0

    • -

  • 0

    • -

  • 0

    • -

  • 0

    • -

-
    -

  • 1

    • -

  • 1

    • -

  • 1

    • -

  • 1

    • -

  • 1

    • -

  • 1

    • -

-
    -

  • 50

    • -

  • 100

    • -

  • 200

    • -

  • 300

    • -

  • 315

    • -

  • 320

    • -

-
    -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

-
    -

  • 5

    • -

  • 10

    • -

  • 19

    • -

  • 32

    • -

  • 30

    • -

  • 37

    • -

-
    -

  • 17

    • -

  • 17

    • -

  • 17

    • -

  • 15

    • -

  • 17

    • -

  • 15

    • -

Publishing metadata contains connection groups

Number of connection groups

-
    -

  • 10

    • -

  • 50

    • -

  • 100

    • -

  • 150

    • -

  • 300

    • -

  • 400

    • -

-
    -

  • 1

    • -

  • 1

    • -

  • 1

    • -

  • 1

    • -

  • 1

    • -

  • 1

    • -

-
    -

  • 100

    • -

  • 100

    • -

  • 100

    • -

  • 100

    • -

  • 100

    • -

  • 100

    • -

-
    -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

-
    -

  • 10

    • -

  • 11

    • -

  • 11

    • -

  • 16

    • -

  • 22

    • -

  • 25

    • -

-
    -

  • 17

    • -

  • 19

    • -

  • 22

    • -

  • 19

    • -

  • 20

    • -

  • 20

    • -

Publishing metadata contains access groups

Number of access groups

-
    -

  • 0

    • -

  • 0

    • -

  • 0

    • -

  • 0

    • -

-
    -

  • 1

    • -

  • 10

    • -

  • 20

    • -

  • 40

    • -

-
    -

  • 100

    • -

  • 100

    • -

  • 100

    • -

  • 100

    • -

-
    -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

-
    -

  • 10

    • -

  • 43

    • -

  • 153

    • -

  • 535

    • -

-
    -

  • 17

    • -

  • 26

    • -

  • 24

    • -

  • 24

    • -
+|Publishing servers contact management server for publishing metadata at same time|Number of publishing servers.|0
0
0
0
0
0|1
1
1
1
1
1|50
100
200
300
315
320|LAN|5
10
19
32
30
37|17
17
17
15
17
15| +|Publishing metadata contains connection groups|Number of connection groups|10
20
100
150
300
400|1
1
1
1
1
1|100
100
100
100
100
100|LAN|10
11
11
16
22
25|17
19
22
19
20
20| +|Publishing metadata contains access groups|Number of access groups|0
0
0
0|1
10
20
40|100
100
100
100|LAN|10
43
153
535|17
26
24
24| The CPU utilization of the computer running the management server is around 25% irrespective of the number of publishing servers targeting it. The Microsoft SQL Server database transactions/sec, batch requests/sec and user connections are identical irrespective of the number of publishing servers. For example, transactions/sec is approximately 30, batch requests approximately 200, and user connects approximately six. Using a geographically distributed deployment, where the management server and publishing servers utilize a slow link network between them, the round-trip response time on the publishing servers is within acceptable time limits (<5 seconds), even for 100 simultaneous requests on a single management server. -|Scenario|Variation|Number of connection groups|Number of access groups|Number of publishing servers|Network connection type|Round-trip response time (seconds)|Management server CPU utilization| +|Scenario|Variation|Number of connection groups|Number of access groups|Number of publishing servers|Network connection type|Round-trip response time (seconds)|Management server CPU utilization (in %)| |---|---|---|---|---|---|---|---| |Network connection between the publishing server and management server|1.5 Mbps Slow link Network|0
0|1
1|50
100|1.5 Mbps Cable DSL|4
5|1
2| |Network connection between the publishing server and management server|LAN/WiFi Network|0
0|1
1|100
200|WiFi|11
20|15
17| - ---------- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
ScenarioVariationNumber of connection groupsNumber of access groupsNumber of publishing serversNetwork connection type publishing server/management serverRound trip response time on the publishing server (in seconds)CPU utilization on management server

Network connection between the publishing server and management server

1.5 Mbps slow link network

-
    -

  • 0

    • -

  • 0

    • -

-
    -

  • 1

    • -

  • 1

    • -

-
    -

  • 50

    • -

  • 100

    • -

-
    -

  • 1.5 Mbps Cable DSL

    • -

  • 1.5 Mbps Cable DSL

    • -

-
    -

  • 4

    • -

  • 5

    • -

-
    -

  • 1

    • -

  • 2

    • -

Network connection between the publishing server and management server

LAN/WiFi network

-
    -

  • 0

    • -

  • 0

    • -

-
    -

  • 1

    • -

  • 1

    • -

-
    -

  • 100

    • -

  • 200

    • -

-
    -

  • Wifi

    • -

  • Wifi

    • -

-
    -

  • 11

    • -

  • 20

    • -

-
    -

  • 15

    • -

  • 17

    • -
- Whether the management server and publishing servers are connected over a slow link network, or a high speed network, the management server can handle approximately 15,000 package refresh requests in 30 minutes. ## App-V Reporting Server capacity planning recommendations @@ -434,60 +108,16 @@ App-V clients send reporting data to the reporting server. The reporting server |Scenario|Summary| |---|---| |Multiple App-V clients send reporting information to the reporting server simultaneously.|Round-trip response time from the reporting server is 2.6 seconds for 500 clients. Round-trip response time from the reporting server is 5.65 seconds for 1000 clients. Round-trip response time increases linearly depending on number of clients.| -|Requests per second processed by the reporting server.|A single reporting server and a single database, can process a maximum of 139 requests per second. The average is 121 requests/second. Using two reporting servers reporting to the same Microsoft SQL Server database, the average requests/second,like a single reporting server, is about 127, with a max of 278 requests/second. A single reporting server can process 500 concurrent/active connections. A single reporting server can process a maximum 1,500 concurrent connections.| +|Requests per second processed by the reporting server.|A single reporting server and a single database, can process a maximum of 139 requests per second. The average is 121 requests/second. Using two reporting servers reporting to the same Microsoft SQL Server database, the average requests/second, like a single reporting server, is about 127, with a max of 278 requests/second. A single reporting server can process 500 concurrent/active connections. A single reporting server can process a maximum 1,500 concurrent connections.| |Reporting database.|Lock contention on the computer running Microsoft SQL Server is the limiting factor for requests/second. Throughput and response time are independent of database size.| - ---- - - - - - - - - - - - - - - - - - - - - -
ScenarioSummary

Multiple App-V clients send reporting information to the reporting server simultaneously.

-
    -

  • Round-trip response time from the reporting server is 2.6 seconds for 500 clients.

    • -

  • Round-trip response time from the reporting server is 5.65 seconds for 1000 clients.

    • -

  • Round-trip response time increases linearly depending on number of clients.

    • -

Requests per second processed by the reporting server.

-

-
    -

  • A single reporting server and a single database, can process a maximum of 139 requests per second. The average is 121 requests/second.

    • -

  • Using two reporting servers reporting to the same Microsoft SQL Server database, the average requests/second,like a single reporting server, is about 127, with a max of 278 requests/second.

    • -

  • A single reporting server can process 500 concurrent/active connections.

    • -

  • A single reporting server can process a maximum 1500 concurrent connections.

    • -

Reporting database.

-

-
    -

  • Lock contention on the computer running Microsoft SQL Server is the limiting factor for requests/second.

    • -

  • Throughput and response time are independent of database size.

    • -
- ### Calculating random delay The random delay specifies the maximum delay (in minutes) for data to be sent to the reporting server. When the scheduled task is started, the client generates a random delay between **0** and **ReportingRandomDelay** and will wait the specified duration before sending data. -Random delay = 4 \* number of clients / average requests per second. (CHECK) +Random delay = 4 * number of clients/average requests per second. -Example: For 500 clients, with 120 requests per second, the Random delay is, 4 \* 500 / 120 = about 17 minutes. (CHECK) +Example: Random delay for 500 clients with 120 requests per second is 4 * 500/120 = about 17 minutes. ## App-V publishing server capacity planning recommendations @@ -495,7 +125,6 @@ Computers running the App-V client connect to the App-V publishing server to sen >[!IMPORTANT] >The following list displays the main factors to consider when setting up the App-V publishing server: - * The number of clients connecting simultaneously to a single publishing server. * The number of packages in each refresh. * The available network bandwidth in your environment between the client and the App-V publishing server. @@ -506,47 +135,6 @@ Computers running the App-V client connect to the App-V publishing server to sen |Number of packages in each refresh.|Increasing number of packages will increase response time by about 40% (up to 1,000 packages).| |Network between the App-V client and the publishing server.|Across a slow network (1.5 Mbps bandwidth), there is a 97% increase in response time compared to LAN (up to 1,000 users).| - ---- - - - - - - - - - - - - - - - - - - - - -
ScenarioSummary

Multiple App-V clients connect to a single publishing server simultaneously.

-
    -

  • A publishing server running dual core processors can respond to at most 5000 clients requesting a refresh simultaneously.

    • -

  • For 5,000–10,000 clients, the publishing server requires a minimum quad core.

    • -

  • For 10,000–20,000 clients, the publishing server should have dual quad cores for more efficient response times.

    • -

  • A publishing server with a quad core can refresh up to 10,000 packages within three seconds. (Supports 10,000 simultaneous clients.)

    • -

Number of packages in each refresh.

-

-
    -

  • Increasing number of packages will increase response time by about 40% (up to 1,000 packages).

    • -

Network between the App-V client and the publishing server.

-

-
    -

  • Across a slow network (1.5 Mbps bandwidth), there is a 97% increase in response time compared to LAN (up to 1,000 users).

    • -
- >[!NOTE] >The publishing server CPU usage is always high during the time interval when it must process simultaneous requests (>90% in most cases). The publishing server can handle about 1,500 client requests in one second. @@ -556,153 +144,12 @@ Computers running the App-V client connect to the App-V publishing server to sen |Multiple packages in each refresh.|Number of packages|1,000
1,000|500
1,000|Quad Core|LAN|2
3|92
91| |Network between client and publishing server.|1.5 Mbps Slow link network|100
500
1,000|120
120
120|Quad Core|1.5 Mbps intra-continental network|3
10 (0.2% failure rate)
7 (1% failure rate)|| - ---------- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
ScenarioVariationNumber of App-V clientsNumber of packagesProcessor configuration on the publishing serverNetwork connection type publishing server/App-V clientRound trip time on the App-V client (in seconds)CPU utilization on publishing server (in %)

App-V client sends publishing refresh request & receives response, each request containing 120 packages

Number of clients

-
    -

  • 100

    • -

  • 1,000

    • -

  • 5,000

    • -

  • 10,000

    • -

-
    -

  • 120

    • -

  • 120

    • -

  • 120

    • -

  • 120

    • -

-
    -

  • Dual core

    • -

  • Dual core

    • -

  • Quad core

    • -

  • Quad core

    • -

-
    -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

-
    -

  • 1

    • -

  • 2

    • -

  • 2

    • -

  • 3

    • -

-
    -

  • 100

    • -

  • 99

    • -

  • 89

    • -

  • 77

    • -

Multiple packages in each refresh

Number of packages

-
    -

  • 1,000

    • -

  • 1,000

    • -

-
    -

  • 500

    • -

  • 1,000

    • -

-
    -

  • Quad core

    • -

  • Quad core

    • -

-
    -

  • LAN

    • -

  • LAN

    • -

-
    -

  • 2

    • -

  • 3

    • -

-
    -

  • 92

    • -

  • 91

    • -

Network between client and publishing server

1.5 Mbps slow link network

-
    -

  • 100

    • -

  • 500

    • -

  • 1,000

    • -

-
    -

  • 120

    • -

  • 120

    • -

  • 120

    • -

-
    -

  • Quad core

    • -

  • Quad core

    • -

  • Quad core

    • -

-
    -

  • 1.5 Mbps intra-continental network

    • -

-
    -

  • 3

    • -

  • 10 (with 0.2% failure rate)

    • -

  • 17 (with 1% failure rate)

    • -

- ## App-V streaming capacity planning recommendations Computers running the App-V client stream the virtual application package from the streaming server. Round trip response time is measured on the computer running the App-V client, and is the time taken to stream the entire package. >[!IMPORTANT] >The following list identifies the main factors to consider when setting up the App-V streaming server: - * The number of clients streaming application packages simultaneously from a single streaming server. * The size of the package being streamed. * The available network bandwidth in your environment between the client and the streaming server. @@ -713,44 +160,6 @@ Computers running the App-V client stream the virtual application package from t |Size of the package being streamed.|The package size has a significant impact on the streaming/download time only for larger packages with a size of about 1 GB. For package sizes ranging from 3 MB to 100 MB, the streaming time ranges from 20 seconds to 100 seconds, with 100 simultaneous clients.| |Network between the App-V client and the streaming server.|Across a slow network (1.5 Mbps bandwidth), there is a 70–80% increase in response time compared to LAN (up to 100 users).| - ---- - - - - - - - - - - - - - - - - - - - - -
ScenarioSummary

Multiple App-V clients stream applications from a single streaming server simultaneously.

-
    -

  • If the number of clients simultaneously streaming from the same server increases, there is a linear relationship with the package download/streaming time.

    • -

Size of the package being streamed.

-

-
    -

  • The package size has a significant impact on the streaming/download time only for larger packages with a size of about 1 GB. For package sizes ranging from 3 MB to 100 MB, the streaming time ranges from 20 seconds to 100 seconds, with 100 simultaneous clients.

    • -

Network between the App-V client and the streaming server.

-

-
    -

  • Across a slow network (1.5 Mbps bandwidth), there is a 70–80% increase in response time compared to LAN (up to 100 users).

    • -
- The following table displays sample values for each of the factors in the previous list: |Scenario|Variation|Number of App-V clients|Size of each package|Network connection type|Round-trip time on the App-V client (in seconds)| @@ -759,131 +168,6 @@ The following table displays sample values for each of the factors in the previo |Size of each package being streamed.|Size of each package.|100
200
100
200|21 MB
21 MB
109 MB
109 MB|LAN|33
83
100
160| |Network connection between client and App-V streaming server.|1.5 Mbps Slow link network.|100
100|3.5 MB
5 MB|1.5 Mbps intra-continental network|102
121| - -------- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
ScenarioVariationNumber of App-V clientsSize of each packageNetwork connection type streaming server/App-V clientRound trip time on the App-V client (in seconds)

Multiple App-V clients streaming virtual application packages from a streaming server.

Number of clients.

-
    -

  • 100

    • -

  • 200

    • -

  • 1,000

    • -

    • -

  • 100

    • -

  • 200

    • -

  • 1,000

    • -

-
    -

  • 3.5 MB

    • -

  • 3.5 MB

    • -

  • 3.5 MB

    • -

    • -

  • 5 MB

    • -

  • 5 MB

    • -

  • 5 MB

    • -

-
    -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

    • -

  • LAN

    • -

  • LAN

    • -

  • LAN

    • -

-
    -

  • 29

    • -

  • 39

    • -

  • 391

    • -

    • -

  • 35

    • -

  • 68

    • -

  • 461

    • -

Size of each package being streamed.

Size of each package.

-
    -

  • 100

    • -

  • 200

    • -

    • -

  • 100

    • -

  • 200

    • -

-
    -

  • 21 MB

    • -

  • 21 MB

    • -

    • -

  • 109

    • -

  • 109

    • -

-
    -

  • LAN

    • -

  • LAN

    • -

    • -

  • LAN

    • -

  • LAN

    • -

-

33

-

83

-

-

100

-

160

Network connection between client and App-V streaming server.

1.5 Mbps slow link network.

-
    -

  • 100

    • -

    • -

  • 100

    • -

-
    -

  • 3.5 MB

    • -

    • -

  • 5 MB

    • -

-
    -

  • 1.5 Mbps intra-continental network

    • -

-

102

-

-

121

- Each App-V streaming server should be able to handle a minimum of 200 clients concurrently streaming virtualized applications. >[!NOTE]