Okay, so in this blog posting I want to continue covering a few more views in Edgesight that I like to run ad hoc queries against. Today’s view is called vw_es_archive_application_network_performance. This view provides information network delay, server delay, xenapp server, process name and downstream hosts that your XenApp servers communicate with. I have used this table to check delays of the executables such as winlogon.exe to check delay between this process and our domain controllers. I will cover checking delays by process name, xen_app server and downstream host.
The first part will be to demonstrate how to find Network and Server delay of specific downstream hosts as well as how to measure the average XenAPP Servers delay. Then in the second part I want to answer one of the questions from the first posting.
Down Stream Delay:
I actually got to present on Edgesight during Synergy 2008 and one of the key points that I tried to drive home is how Edgesight helps you with the never ending B.S. Witch hunts that always seem to occur when someone’s application is “running slow on Citrix”. I would say that less than 30 % of what I actually investigate ends up being an actual XenAPP issue. I will go over a few ad hoc queries that will give you the average delay of your down stream hosts and will give you the average delay experienced by each XenAPP Server allowing you to see if you have a specific XenAPP box that may be having some issues.
The first ad hoc query has to do with downstream hosts, this will return the downstream host and the Network/Server delay. I have set this query to filter any downstream host that does not have at least 100 records and a server delay of at least 300 miliseconds. You can edit/remove the “Having” clause to suit your environment.
select distinct hostname, sum(network_delay_sum)/sum(record_count) as “Network Delay”, sum(server_delay_sum)/sum(record_count) as “Server Delay”
from vw_es_archive_application_network_performance
group by hostname
having sum(record_count) > 100
and sum(server_delay_sum)/sum(record_count) > 300
order by sum(server_delay_sum)/sum(record_count) desc
In English: “Give me the Network and Server delay of every downstream host that has at least 100 records (packets?) and a server latency of at least 300ms”
XenAPP Server Delay:
It is a good idea to monitor your XenAPP Server delay, this will tell you if there is a particular XenAPP Server that is having a layer 1 or layer 2 issue. This is a quick query that will show you the average delay of your XenAPP Servers.
select distinct machine_name, sum(network_delay_sum)/sum(record_count) as “Network Delay”, sum(server_delay_sum)/sum(record_count) as “Server Delay”
from vw_es_archive_application_network_performance
group by machine_name
order by sum(server_delay_sum)/sum(record_count) desc
Note: You will also see “Edgesight for Endpoints” client data in this table as well.
Executable Delay:
This query shows the delay associated individual executables. You may check outlook.exe to see if you have a delay in a downstream Exchange server or, in my case, check winlogon.exe for delays to domain controllers.
select distinct exe_name, sum(network_delay_sum)/sum(record_count) as “Network Delay”, sum(server_delay_sum)/sum(record_count) as “Server Delay”
from vw_es_archive_application_network_performance
group by exe_name
order by sum(server_delay_sum)/sum(record_count) desc
Session Statistics:
Last week I got a a question about session counts and I wanted to answer it in this post, here was the question:
”I’m looking for a custom report showing the application usage (Published Apps, not processes) on a hourly, daily and monthly base and a custom report showing the concurrent sessions on a hourly, daily and monthly base.”
The view I used for this was vw_ctrx_archive_client_start_perf declare @end varchar
declare @today datetime
declare @app varchar
set @today = convert(varchar,getdate(),111)
set @begin = ‘00′
set @end = ‘23′
set @app = ‘%Outlook%’
select convert(varchar(2),dateadd(hh,-4,time_stamp), 108)+’:00′ as “Time”, count(distinct sessid)
from vw_ctrx_archive_client_start_perf
where convert(varchar(10),dateadd(hh,-4,time_stamp), 111) = @today-1
and published_application like ‘%’+@app+’%’
group by convert(varchar(2),dateadd(hh,-4,time_stamp), 108)+’:00′
order by convert(varchar(2),dateadd(hh,-4,time_stamp), 108)+’:00′
In English: Give me every application on an hourly basis for a specific application. On this report substitute %APPNAME% for whichever app you want to see. Note that this is an hourly report so the time format is set to 108.
Daily Application Usage:
In the same view I change the query above just a little to accommodate a query by day.
declare @begin varchar
declare @end varchar
declare @today datetime
declare @app varchar
set @today = convert(varchar,getdate(),111)
set @app = ‘%Outlook%’
select convert(varchar(10),dateadd(hh,-4,time_stamp), 111) as “Date”, count(distinct sessid)
from vw_ctrx_archive_client_start_perf
where convert(varchar(10),dateadd(hh,-4,time_stamp), 111) > @today-30
and published_application like ‘%’+@app+’%’
group by convert(varchar(10),dateadd(hh,-4,time_stamp), 111)
order by convert(varchar(10),dateadd(hh,-4,time_stamp), 111)
Monthly Application Usage:
Depending on how long you have your retention set (min is 30 days) this query may or may not work for you but this is the number of unique sessions per application for a month.
declare @begin varchar
declare @end varchar
declare @today datetime
declare @app varchar
set @today = convert(varchar,getdate(),111)
set @app = ‘%Outlook%’
select convert(varchar(7),dateadd(hh,-4,time_stamp), 111) as “Date”, count(distinct sessid)
from vw_ctrx_archive_client_start_perf
where convert(varchar(10),dateadd(hh,-4,time_stamp), 111) > @today-30
and published_application like ‘%’+@app+’%’
group by convert(varchar(7),dateadd(hh,-4,time_stamp), 111)
order by convert(varchar(7),dateadd(hh,-4,time_stamp), 111)
Application Matrix:
SQL Server Reporting Services will let you create a matrix, these two queries are for daily and monthly which will let you sort as follows:
| Date 1 | Date2 | Date3 | Date4 | Date5 | |
| Outlook | Count1 | Count2 | Count3 | Count4 | Count5 |
| Word | Count1 | Count2 | Count3 | Count4 | Count5 |
| Oracle Financials | Count1 | Count2 | Count3 | Count4 | Count5 |
| Statistical APP | Count1 | Count2 | Count3 | Count4 | Count5 |
| Custom APP-A | Count1 | Count2 | Count3 | Count4 | Count5 |
This has been the report method that has made my management the happiest so I use the Matrix tool with SSRS as often as possible. Remember, if you have Edgesight, you have SSRS and setting up reports is no harder than an Access Database.
Here are the queries
First The Daily Matrix:
declare @begin varchar
declare @end varchar
declare @today datetime
declare @app varchar
set @today = convert(varchar,getdate(),111)
select convert(varchar(10),dateadd(hh,-4,time_stamp), 111) as “Date”, published_application, count(distinct sessid)
from vw_ctrx_archive_client_start_perf
where convert(varchar(10),dateadd(hh,-4,time_stamp), 111) > @today-30
group by convert(varchar(10),dateadd(hh,-4,time_stamp), 111), published_application
order by convert(varchar(10),dateadd(hh,-4,time_stamp), 111), count(distinct sessid) desc
Then the Monthly Matrix:
declare @today datetime
set @today = convert(varchar,getdate(),111)
select convert(varchar(7),dateadd(hh,-4,time_stamp), 111) as “Date”, published_application, count(distinct sessid)
from vw_ctrx_archive_client_start_perf
where convert(varchar(10),dateadd(hh,-4,time_stamp), 111) > @today-30
group by convert(varchar(7),dateadd(hh,-4,time_stamp), 111), published_application
order by convert(varchar(7),dateadd(hh,-4,time_stamp), 111), count(distinct sessid) desc
Concurrent Session Statistics:
A colleague of mine, Alain Assaf, set up a system that gives you this info every five minutes and is almost in real time, go to wagthereal.wordpress.com to see it. Keep in mind that Edgesight is not real time data so if you set up a private dashboard for it, you may have to wait for it to refresh.
The vw_ctrx_archive_client_start_perf view appears to give us only start times of specific published applications. Perhaps the most used view of any of my reports is vw_ctrx_archive_ica_roundtrip_perf. For this set of queries, I will count concurrent sessions but I will also go into ICA Delay’s for clients in my last post on Edgesight Under the Hood:
I will try to answer the users question on concurrent sessions with three pretty basic queries for hourly, daily and monthly usage:
Hourly Users:
declare @begin varchar
declare @end varchar
declare @today datetime
declare @app varchar
set @today = convert(varchar,getdate(),111)
set @begin = ‘00′
set @end = ‘23′
select convert(varchar(2),dateadd(hh,-4,time_stamp), 108)+’:00′ as “Time”, count(distinct [user])
from vw_ctrx_archive_ica_roundtrip_perf
where convert(varchar(10),dateadd(hh,-4,time_stamp), 111) = @today-3
group by convert(varchar(2),dateadd(hh,-4,time_stamp), 108)+’:00′
order by convert(varchar(2),dateadd(hh,-4,time_stamp), 108)+’:00′
Daily Users:
declare @begin varchar
declare @end varchar
declare @today datetime
declare @app varchar
set @today = convert(varchar,getdate(),111)
select convert(varchar(10),dateadd(hh,-4,time_stamp), 111) as “Date”, count(distinct [user])
from vw_ctrx_archive_ica_roundtrip_perf
where convert(varchar(10),dateadd(hh,-4,time_stamp), 111) > @today-30
group by convert(varchar(10),dateadd(hh,-4,time_stamp), 111)
order by convert(varchar(10),dateadd(hh,-4,time_stamp), 111)
Monthly Users:
declare @begin varchar
declare @end varchar
declare @today datetime
declare @app varchar
set @today = convert(varchar,getdate(),111)
select convert(varchar(7),dateadd(hh,-4,time_stamp), 111) as “Date”, count(distinct [user])
from vw_ctrx_archive_ica_roundtrip_perf
where convert(varchar(10),dateadd(hh,-4,time_stamp), 111) > @today-30
group by convert(varchar(7),dateadd(hh,-4,time_stamp), 111)
order by convert(varchar(7),dateadd(hh,-4,time_stamp), 111)
Conclusion:
For the most part, I have vetted all of these queries, you may get varying results, if so, check for payload errors, licensing, etc. I would really like to see some better documentation on the data model, most of these were basically done by running the query and checking it against the EdgeSight canned reports to see if my SWAG about how they did their calculations was correct. All of the queries I ran here I checked and looked to be accurate. If you are going to bet the farm on any of these queries to the brass in your organization, vet my numbers….
My next post will deal with ICA latency and delay issues for individual users and servers.
Thanks for reading!
John
