Impact of Latency on Project Professional 2007



Please find below a study I did for a customer to demonstrate the advantage of the new EPM 2007 architecture and specifically how Project Professional 2007 communicates to the application server.


DISCLAIMER: the times presented below highly depend on the EPM Architecture, the project plan data, as the network between the test machine and the servers, as well as the testing methodology. There is thus no guarantee on the response time in your environment!


 


EXECUTIVE SUMMARY


Contoso is a worldwide company with many branch offices and is considering a migration from Project Server 2003 to Project Server 2007 but requires reassurance that the newer version of EPM would handle high latency environment when using Project Professional 2007. Contoso uses a Citrix farm today to address the latency issues of Project Professional 2003.


The results showed that Project Professional 2007 runs fine under a high latency environment thus Contoso can migrate from EPM 2003 to EPM 2007 and not use a Citrix farm anymore to run Project Professional.


TEST ENVIRONMENT


EPM 2007 Architecture


Project Server 2007 was installed on a single server (including SQL 2005) with the following specifications:


Hardware


¾  Computer modem:     Dell Power Edge 2850


¾  Processor:                Dual proc Xeon 3.6 Ghz


¾  System memory:       4 Go, FSB 800 Mhz


¾  Hard drive:               36 Go Ultra Scsi 320 15000 rpm


Software


¾  Windows 2003 R2, Enterprise Edition US + SP1


¾  SQL Server 2005, Enterprise Edition US + SP1


¾  Project Server 2007, x86 EN


Test Framework


The EPM Test Framework (EPMTF) was used to automate the testing process of the Project Professional 2007.


The EPMTF ran on a Dell Latitude D600 laptop with 2 Go RAM; Pentium 1,7GHZ & 7200 rpm hard drive running Windows Server 2003 + SP1.


No latency simulation tool were used and instead a true 120 ms latency was generated by accessing an EPM 2007 server located at the MS France lab and accessing it from home via a VPN connection.


Test Data


¾  Contoso gave us a generic project plan used to track “business as usual” activities.


¾  This initial plan contained the following:


¾  440 tasks


¾  10 projects Enterprise Custom Fields


¾  10 resource Enterprise Custom Fields


¾  5 task Enterprise Custom Fields


¾  5 resources


¾  1 to 5 resource assignment per task


¾  No baseline


To simulate the impact of project size we multiplied the total number of tasks for each run as shown below, total number of tasks per run:


1.      444


2.      877


3.      1304


4.      1733


5.      2160


Use Case


Only one use case was tested, it included the following actions:


1.      Start Project Professional 2007 and open project plan


2.      Add one task at the end of the project plan


3.      Save project plan


4.      Publish project plan


5.      Close project plan


TEST RESULTS


Response Time


Average response time in milliseconds for five runs for each action and each project size with 120ms latency:


















































Total Number of Project Tasks (results in ms)


 


444


877


1304


1733


2160


Open


7 415


7 283


7 645


7 782


8 533


AddTask


78


188


313


570


464


Save


1 883


2 438


3 003


3 426


5 486


Publish


5 672


5 993


5 628


5 534


3 057


Close


1 781


1 289


1 086


1 138


1 157


The following observations can be made from the results above:


¾  The time to Open a plan “stays” under 10 seconds which is acceptable.


¾  Close & Publish actions are not impacted by the project size (number of tasks).


Active Cache


The following table shows the percentage gain (compared to the table above) obtained by using the Active Cache when performing each action:


















































Total Number of Project Tasks


444


877


1304


1733


2160


Open


33%


79%


127%


188%


255%


AddTask


-35%


-58%


7%


-64%


-15%


Save


6%


-2%


-2%


0%


-30%


Publish


-5%


-13%


-9%


-11%


68%


Close


-54%


7%


127%


111%


293%


The following observations can be made from the results above:


¾  As expected the AC has mostly an impact when opening the project plan.


¾  AC does generate significant time improvement as the project plan increases in size.


Raw results (in milliseconds):































































































Total Number of Project Tasks (results in ms)


 


 


444


877


1304


1733


2160


Open


Active Cache


7,415


7,283


7,645


7,782


8,533


 


No Active Cache


9,852


13,052


17,372


22,425


30,293


AddTask


Active Cache


78


188


313


570


464


 


No Active Cache


51


79


336


207


394


Save


Active Cache


1,883


2,438


3,003


3,426


5,486


 


No Active Cache


2,003


2,377


2,952


3,440


3,827


Publish


Active Cache


5,672


5,993


5,628


5,534


3,057


 


No Active Cache


5,407


5,200


5,132


4,915


5,120


Close


Active Cache


1,781


1,289


1,086


1,138


1,157


 


No Active Cache


818


1,380


2,465


2,398


4,541


 


CONCLUSION


Opening/Publishing/Saving actions in Project Professional 2007 are well under 10 seconds with 120ms latency; the responses time are thus acceptable.


The new Project Server 2007 architecture introduces significant improvement in performance when using Project Professional in a high latency environment. The performance improvement is mainly due to the fact the data access layer was rewritten to use web services to access the database instead of ODBC connections (EPM 2003). 


Active Cache does provide further performance improvement when opening a large project plan after it has been open once before (cached copy of project plan is created the very first time a plan is open).


In theory Contoso does not need to setup a Citrix farm to use Project Professional 2007 from a performance point a view but might still require it if the client desktop does not meet the Office 2007 minimum requirements of XP SP2 for the OS.

Comments (1)

  1. Please find below an inventory of tools that can be used to perform the following tests: Capacity Planning

Skip to main content