JetStress (WR2.1): 9000 log buffers became the default for Exchange 2003.

To correctly set the JET_paramLogBuffers value, JetStress uses the following information For Exchange 2000, Exchange 2000 SP1, Exchange 2000 SP2, or Exchange 2000 SP3, set the value to 512. For Exchange 2003, Exchange 2003 SP1, or Exchange 2003 SP2, set the value to 9000. Reference: ESE log buffers that are set too low can cause the Information Store…

0

JetStress (WR2.2): Unhandled Exception: Cannot read Instance : F:MDBSG1

I have come across System.InvalidOperationException: Cannot read Instance : F:\MDB\SG1. From memory dump analysis with the .NET perf. team, the main cause was either one of these: Some logical disks are not fully ready to use in the Windows PDH library. Something went wrong in the .NET perf. library (maybe thread-safety issues). See the following example: 12/22/2005 9:32:10 AM — Start tuning…

0

Jetstress: Performance characteristic is more realistic when 80% of the disk capacity is utilized.

Some tests showed that using 70% or 80% of the disk capacity can show the realistic performance characteristic. The achieved IO is up to 100% more than what you can really get when 80% of the disk is utilized in terms of its capacity. Refer to a TechNet traing session slide: http://download.microsoft.com/documents/uk/technet/learning/downloads/evenings/Exchange_2003_Storage_Design_Brad_Carter_14th_July.ppt This posting is…

0

JetStress (WR2.1): There has been confusion about sizing test database.

JetStress (WR 1.2: 06.05.7529.0) — Calculation is like the following: Planned Data: 750 users * 200 MB = 150 GB. Initial Data: 150 GB * 75% = 112.5 GB. Initial EDB: 112.5 GB * 1.67% = approximately 188 GB — 112.5 GB mailbox data + 66% overhead. Required space (which is double the initial data)…

0

JetStress (WR2.1) uses Database Page Fault Stall Per Sec.

JetStress (WR2.1) uses Database Page Fault Stalls Per Sec instead of Database disk write latency. It’s mainly because it could handle SRDF/sync environment. In SRDF/sync environment, database disk write latency can go over 20 ms on average, but it is not necessarily causing users to experience performance problems. However, in a non-replicated environment, if the database write…

0

JetStress: Database Fault Stalls/sec should be zero on production servers.

http://www.microsoft.com/technet/prodtechnol/exchange/guides/TrblshtE2k3Perf/5bcdd349-dcc6-43eb-9dc3-54175f7061ad.mspx Exchange servers should generally have database write latencies under 20 ms, with spikes (maximum values) under 50 ms. However, it is not always possible to keep write latencies in this range when synchronous replication is in use. Database write latency issues often do not become apparent to the end user until the database cache…

0

JetStress: Differences between performance and stress test runs

There are some differences between two test run types (performance and stress): Tuning phase uses different disk latency thresholds. See the JetStress documentation for the details.(the next web release of JetStressUI will allow users to edit the default disk latency thresholds in JS_Config.xml. But, this should be used with caution since high thresholds can break down the disk…

0

Tip: Office Live Meeting can help live debugging.

Easy Assist https://www.livemeeting.com/cc/easyassist Schedule a meeting or meet now at https://www.livemeeting.com/cc/microsoft or (https://lmportal/ to setup acccout and password) Send E-mail Invite (from Office Live Meeting console) or you can send the attendee URL to your attendees. You can find the attendee URL at https://www.livemeeting.com/cc/microsoft/manageMeetings (shows all meetings). Here is an example of attendee URL: https://www.livemeeting.com/cc/microsoft/join?id=hmlee&role=attend&pw=jp2%2C%7E%3DPXr

0

JetStress (WR2.1) has updated performance counter thresholds.

Performance counter thresholds are updated. But, there are some differences from Exchange 2003 Performance Troubleshooting Guide for some reasons. Database Disks: Average Disk sec/Read (or I/O Database Reads Average Latency): The average should be below 20 ms. Spikes (maximum values) should not be higher than 50 ms. Transactional Log Disks: Average Disk sec/Read (or I/O Log…

1