Continuous calls made by the devices to Exchange can cause enough "I/O thrash" to bring "an Exchange server to a grinding halt," said Steve Lewis, co-founder and CEO at Terneros Inc., which makes the Application Continuity Appliance (ACA) for Microsoft Exchange, a high-availability product.
Teneros discovered this problem when it installed an ACA rated for 250 Exchange users at a customer site with 65 employees -- all of them BlackBerry users. Because one BlackBerry user is equal to approximately 4.5 standard Exchange users, the ACA was "struggling," Lewis said. Similar performance problems happen when users have very large Exchange databases.
Teneros engineers fixed the problem by adding more disk spindles to the ACA. "With BlackBerry users, we've found that customers have to double down on their
Even without Blackberrys clogging up the system, Exchange in general requires more disk spindles than users might think, especially in older versions. New enhancements to Exchange 2007, including support for 64-bit operating systems and replication features, could help ease the strain on production systems by increasing the memory from 4 GB to 8 GB.But the system still relies on the original .JET database, considered inefficient by today's standards. The database does sequential reads but random writes, making searching cumbersome -- and the bigger the disk spindle, the longer the latency during this process.
Bill Augustadt, chief technology officer of Application Managed Services, ACS, Inc., which manages IT for blue-chip companies, like Disney, NBC and Nike, and stores a total of close to 3 petabytes (PB) of data between its internal IT and the IT it manages for clients, said more spindles is the way to go when managing Exchange -- across the board.
"After 30 milliseconds, a bubble pops up on a user's screen saying 'waiting for network,' " Augustadt said. "The only way to avoid that is to spread the Exchange data over more spindles."
This article originally appeared on SearchStorage.com