We are planning to use Jetstress to verify the pre UAT/performance of our hardware configuration prior to installing Exchange and LoadGen for the formal UAT/Performance test of the Exchange configuration. We need to put together a ROM for the amount of time to performance test our configuration. We are configuring a system to handle 190,000 users, 5% Heavy - 1 GB Mail Box - avg 60 messages per day, peak 300 messages per day, 10% Medium - 500 MB Mail Box - avg 37 messages per day, peak 128 messages per day, 85% Light - 100 MB Mail Box - avg 43 messages per day, peak 188 messages per day, about 7% also use blackberry to access their exchange mail. I have not, and will not have, the time or opportunity to build a lab to ramp up tests before the ROM is due. If I can get a valid timeframe for testing time from these numbers I would appreciate it, and/or I am hoping someone can give me a way to take this information along with the hardware configuration to get a valid timeframe for our testing. This way I can use that to estimate any changes to our configuration or support my contention that the 10 days as proposed by our senior tech is nowhere near enough time for the performance testing. I know that this may be asking for the crown jewels of information but any information would be greatly appreciated.

If I can get a valid timeframe for testing time from these numbers I would appreciate it... For JetStress: "A full validation should run for at least 8 hours". 2-24 are the outside limits. Source: Exchange 2010 Best Practices (Jagott and Stidley, page 597). I'm assuming Exchange 2010 here, right? No off chance this is for E2K7 for some reason? I have not, and will not have, the time or opportunity to build a lab to ramp up tests before the ROM is due. What will you test this on? Loadgen requires a functional Exchange server and client computer. I have no idea what kind of hardware that kind of load would require. Never worked with anything close to that number of users, heavy load or not.

Yes this is Exchange 2010 and we will be using Jetstress 2010 and Exchange Load Generator (LoadGen) 2010. This is very large company which lives off of it's email system. The requirements are that the solution support 190,000 total users worldwide. Approx. 35.5 Million messages a day with an Avg message size of 1MB and a Max of 30MB. There are quite a few servers and also a few TB of SAN involved. Each server slotted for Exchange will run Jetstress with a portion of the load to exersize the infrstructure before we install Exchange. Jetstress, and all of its remnants, will be removed and Exchange will be installed. After the installation of Exchange we will use LoadGen from a number of client platforms to perform a similar test. Our goal is to ensure that the network and hardware can support the solution and then be able to tune exchange, if needed, to also support it. The servers and SAN are in the production ready environment and, once we have verified the throughput and removed every remnant of the testing, will be used to accept the transition of the existing production environment. I can accept the 8 hours of testing for jetstress but is that 8 hours after it has achieved steady state or an assumption that steady state will take 2 hours and then get 6 hours of steady state. As you can see by our numbers we are talking many large MBXs so I feel that Steady State will not be achieved for quite a few hours. That is why I am asking about Jetstress. As to LoadGen we have the Tier breakdown but I do not know how long to estimate for the testing so we can get user signoff to the requirements. 2 hours will in no way show enough work. The high end number I have seen is 4 months which is also not reasonable.

Be careful about Jetstress testing in a SAN environment, well, anywhere really, not just SAN. Most shared storage solutions use cache at various stages along the line and Jetstress tests often yield utterly pointless results. Also remember what Jetstress does in terms of the working set and the IO profile. With a 2010 environment youre going to treat the results with a pinch of salt and then move on to a more realistic Loadgen testing. A recent conversation a colleague had with a Microsoft field person said that Jetstress was merely to get a representative loading and should not be used for pass/fail purposes. "Herkimer" wrote in message news:c795093f-92d3-404e-8786-aee61ba01639... Yes this is Exchange 2010 and we will be using Jetstress 2010 and Exchange Load Generator (LoadGen) 2010. This is very large company which lives off of it's email system. The requirements are that the solution support 190,000 total users worldwide. Approx. 35.5 Million messages a day with an Avg message size of 1MB and a Max of 30MB. There are quite a few servers and also a few TB of SAN involved. Each server slotted for Exchange will run Jetstress with a portion of the load to exersize the infrstructure before we install Exchange. Jetstress, and all of its remnants, will be removed and Exchange will be installed. After the installation of Exchange we will use LoadGen from a number of client platforms to perform a similar test. Our goal is to ensure that the network and hardware can support the solution and then be able to tune exchange, if needed, to also support it. The servers and SAN are in the production ready environment and, once we have verified the throughput and removed every remnant of the testing, will be used to accept the transition of the existing production environment. I can accept the 8 hours of testing for jetstress but is that 8 hours after it has achieved steady state or an assumption that steady state will take 2 hours and then get 6 hours of steady state. As you can see by our numbers we are talking many large MBXs so I feel that Steady State will not be achieved for quite a few hours. That is why I am asking about Jetstress. As to LoadGen we have the Tier breakdown but I do not know how long to estimate for the testing so we can get user signoff to the requirements. 2 hours will in no way show enough work. The high end number I have seen is 4 months which is also not reasonable. Mark Arnold, Exchange MVP.

Be careful about Jetstress testing in a SAN environment, well, anywhere really, not just SAN. Most shared storage solutions use cache at various stages along the line and Jetstress tests often yield utterly pointless results. Also remember what Jetstress does in terms of the working set and the IO profile. With a 2010 environment youre going to treat the results with a pinch of salt and then move on to a more realistic Loadgen testing. A recent conversation a colleague had with a Microsoft field person said that Jetstress was merely to get a representative loading and should not be used for pass/fail purposes. "Herkimer" wrote in message news:c795093f-92d3-404e-8786-aee61ba01639... Yes this is Exchange 2010 and we will be using Jetstress 2010 and Exchange Load Generator (LoadGen) 2010. This is very large company which lives off of it's email system. The requirements are that the solution support 190,000 total users worldwide. Approx. 35.5 Million messages a day with an Avg message size of 1MB and a Max of 30MB. There are quite a few servers and also a few TB of SAN involved. Each server slotted for Exchange will run Jetstress with a portion of the load to exersize the infrstructure before we install Exchange. Jetstress, and all of its remnants, will be removed and Exchange will be installed. After the installation of Exchange we will use LoadGen from a number of client platforms to perform a similar test. Our goal is to ensure that the network and hardware can support the solution and then be able to tune exchange, if needed, to also support it. The servers and SAN are in the production ready environment and, once we have verified the throughput and removed every remnant of the testing, will be used to accept the transition of the existing production environment. I can accept the 8 hours of testing for jetstress but is that 8 hours after it has achieved steady state or an assumption that steady state will take 2 hours and then get 6 hours of steady state. As you can see by our numbers we are talking many large MBXs so I feel that Steady State will not be achieved for quite a few hours. That is why I am asking about Jetstress. As to LoadGen we have the Tier breakdown but I do not know how long to estimate for the testing so we can get user signoff to the requirements. 2 hours will in no way show enough work. The high end number I have seen is 4 months which is also not reasonable. Mark Arnold, Exchange MVP.

How does your test-environment look like? e.g. how many users with what mailbox size will you test? Don't underestimate the time that JetStress and Loadgen will need to initialize mailboxes! Just fyi, I've recently worked on a PoC for 4500 users with 250MB mailboxes which took (depending on the config) 2-4 days to initialize!!!!