Hello All,
I am in process of deciding harware sizing for setting up new webMethods environment(Integration Server and Broker Server). I need to estimate and propose hardare required(processor type, speed,number of processor, memory etc) to run IS 6.5 and Broker 6.5 on wintel system.
what are the methodology/approach to get the hardare estimation. I have gone through the GEAR docs, but cannot decide anything firmly because i can’t do load testing/stress testing as wMsuggest. On what should i do the stress/loading testing if we dont have a hardware itself(catch 22).
How does consultants genereally suggest hardware requirements ? My integration is very simple. IS would subscirbe mapped canonical documents from broker and insert/update the SQL sever table. This canonical contains only one record with max 40 fields. The transaction volume we expect is around 15K in a day and peak volume can be 2K transaction during load time.
Can some one provide their system details for me to have a benchmark where more then 15 K records are processed daily.
While taking decision on Capacity Planning & Sizing you should consider all your future requirements as well.
There is a very good doument on Gear “Capacity_Planning_&_Sizing_Model”. This is a excel sheet where you just need to fill in your requirements and it will do the rest. You can search this from the gear docs
I have gone through the GEAR docs and but cannot decide anything concrete. Excel sizing sheet is good, but does not recommended hardware. e.g which processor, speed, number of processor etc.
how can i decide this ?
Also wats does this inputs means in sizing sheet.
No of applications - in my case IS would subscribe to mapped documents and insert the records in database.No of end-to-end integrations - does it mean how many packages/interface are deployed ?
It does suggest you on hardware. There is a “CPU Sizing” sheet where you can select which type of hardware you want to use, eg. “Intel Xeon”, “HP PA-RISC”, etc… you can also select the CPU speed and it will suggest you that how many hardware boxes you would need. It also helps you in taking decision on Disk space, Db space and memory. So over all this sheet should give you everything you need.
No of Application: The number of different end system you will be integrating from this IS.
No of end-to-end integrations : This means the number of different integration process you have. which is not actually a package because you may have multiple interfaces in a single package.
I dont see CPU Sizing sheet in the work sheet that i am referring too. Can you pls upload the one u have.
what is integration process ? can you explain a bit more and how does this effect sizing.
whats important is the number of transaction, size of transaction and concurrent transaction, number of adapters and webmethdos components. How can i get sizing based on this.
I am not sure if I can upload the spreadsheet here. The spread sheet says that “Re-distribution of this document is prohibited”
I tried to search Advantage with “Capacity Planning & Sizing Model” and I got following result, which gives the same spreadsheet.
Bookshelf
GEAR 6.5 Capacity Planning & Sizing Model Spreadsheet
You can understand “Integration Process” in many ways. Sometimes the number of canonical I deliver to the end systems becomes “Integration Process”. The intention is to actually find out different type & size of transaction so that the processing speed and memory requirement can be calculated. In an integration you may have different type of transaction like Purchase Requisition, Purchase Order, Delivery Order etc. and each may use different size and may have different number of request.
I would suggest you to refer few of the GEAR docs that give lot of details around it. Eg: “GEAR_6_5_Capacity_Planning”, “wM_6_5_Hardware_JVM_and_Operating_System_Perf_Tuning_Recs”, “GEAR_6_Performance_Tuning_White_Paper” etc.
thanks amar.
Yes i got the sizing sheet that you were refering too. I was looking at the old sizing for 4.6 which didnt had CPU sizing.
few more doubt.
what is “Single Centralized IS - Throttle: IS Throughput” in the Output - Integration Server tab. And why does they have given Bridge 4.6 and IS 6.0 in the left most column.
In the Enterprise Server tab what does “Option 1 Adapter per BO per Application” , “Option 2 Single Adapter per Application”
,“Option 3 Hybrid” means.
Jiten,
Have you already purchased the webMethods components that you will be using? webMethods licensing is generally CPU based so it may dictate some of what you can do. FYI Dual Core CPU’s generally eat 2 CPU licenses(it’s not quite a one to one but close enough in my opinion).
That’s not a ton of volume but that’s just one integration use case. What about future stuff? How will you split out your IS process instances, one per server or multiple per server? Are you running your brokers on a seperate set of servers? How about clustering or load balancing? Did you buy any other webMethods products such as Optimize?
Both webMethods/SoftwareAG and your hardware vendor will work with you on a proof of concept. If they won’t then you should select new ones.
Licensing is not a issue here. We plan to use broker server and Integration server on the same machines. Not having clustering or load balancing right now because of budget contraints, but may go for it mid next year.
We already have this interfaces running in production on HP Unix system (multi CPU). Problem with our current architecuture is Integration server is in One country(Singapore) and legacy is in another country(China). And we do not have broker gateway set up or to say we only have one webMethods IS residing in singapore which fires a SQL statement to legacy systems which are in China. Because of lot network issues I plan to set up another webMethods IS and broker server in China and use publish subscribe model. This IS in China would just subscribe to mapped canonical and insert/update records in legacy.
Sizing sheet came to conclusion that i would need 2 (4-way) 3GHz Intel Xeon boxes.
But still the same question. How can i get the confirmation before buying the hardware or before doing any testing ?
There is no confirmation available for hardware sizing. It’s never an exact science. What you have listed sounds about right for your projected volume. But there are always a ton of variables that can impact performance which are generally unique to an environment. For example 15k transactions is not a lot for your configuration assuming those transactions are spread out over business hours and are not very big. If all 15K come in at the same time its a different picture. That’s what I mean by variable.
If you have throughput requirements such as all 15k transactions must be processed without latency/queuing then that changes your performance profile. If you can tolerate some queuing then thats another. Once you decide what your requirements are for processing the transactions then you move on to performance testing. There isn’t really anyway around that.
If licensing is not an issue then you may want to split the broker server off to it’s own box assuming your network is low latency and reliable. The exception to that would be if your broker is going to use SAN attached storage. It’s not mandatory that you move it to another box but the broker can be a bit IO intensive.
Please search for the “Capacity_Planning_&_Sizing_Model” under the GEAR documentation. This excelsheet is owned by SoftwareAg and I dont think we can share it.