tag:blogger.com,1999:blog-1298619317057174466.post3784058837862468626..comments2009-03-29T21:49:39.647-07:00CSS Labshttp://www.blogger.com/profile/04243355510554494627noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-1298619317057174466.post-74908926083682572102009-03-29T21:49:00.000-07:002009-03-29T21:49:00.000-07:00Great stuff...<BR/>good information...<BR/>Regards,<BR/>http://www.saibposervices.com/Back_office_services.aspx<BR/><A HREF="http://www.saibposervices.com/Back_office_services.aspx" REL="nofollow"><BR/>Payroll services</A>Payroll serviceshttp://www.saibposervices.com/Back_office_services.aspxnoreply@blogger.comtag:blogger.com,1999:blog-1298619317057174466.post-60479606694332701192008-07-09T12:12:00.000-07:002008-07-09T12:12:00.000-07:00If you read the blog carefully, 2 hours is for completing the entier process including delivery of our pay check PDF file. <BR/><BR/>I mentioned abour one hour time period for completing the payroll. <BR/><BR/>The point here is not about the numbers, its about building scalable solution using cloud computing.<BR/><BR/>To answer to your questions,<BR/><BR/>Our local setup has clustered database servers used for various purposes not only for keeping payroll related information.<BR/><BR/>We have two dedicated servers for processing our payroll.<BR/><BR/>As i mentioned in my post, our payroll engine has been built for distributed computing. I believe it uses some messaging infrastrucutre to share their state. We have used only EC2 and S3for running our payroll.<BR/><BR/>We found some latency issues uplaoding directly from india to the AWS S3 servers. But we managed with the help of our US servers.CSS Labshttp://www.blogger.com/profile/04243355510554494627noreply@blogger.comtag:blogger.com,1999:blog-1298619317057174466.post-89381795934874281292008-07-09T09:36:00.000-07:002008-07-09T09:36:00.000-07:00Great stuff... However, I am trying to better understand the ROI here. Initially you say the payroll takes 6 hours to complete.. then you say..<BR/><BR/>"two dedicated servers to process the payroll and we could achieve each payroll cycle in about 3 hours." <BR/><BR/>Later you say it takes two hours then you end with this statement... <BR/><BR/>"enabled us to reduce the processing time considerably (from 5 hours to approximately 1 hour)" <BR/><BR/>I am not trying to be a jerk but I just trying to understand the numbers. Here are a few questions I have: <BR/><BR/>1) What was the original configuration of the dedicated servers (e.g., one MySQL and one Payroll server)?<BR/><BR/>2) When you went to 4 Payroll servers did how did you achieve the parallel processing effect (splitting the data up, SQS, ...)?<BR/><BR/>3) Was their any latency uploading and downloading the S3 data?<BR/><BR/>TIABotchagalupehttp://www.blogger.com/profile/08654543217638378500noreply@blogger.comtag:blogger.com,1999:blog-1298619317057174466.post-57193488537132061822008-07-05T05:57:00.000-07:002008-07-05T05:57:00.000-07:00Good stuff. Thanks for sharing.Anonymousnoreply@blogger.com