Reply
Administrator
Posts: 385
Registered: 02-28-2007
0

Exchange 2007 acceleration results

[ Edited ]
With the release of RiOS 5.0, we have our first performance results for Microsoft Office Outlook 2007 and Exchange Server 2007. RiOS 5.0 is the first release to provide protocol optimization for the Exchange 2007 protocol (MAPI 2K7) at the application layer.

Steelhead appliance gave us the following results:

More than 21x faster email transmissions to multiple recipients in a branch on a T1 link
More than 53x faster email transmissions to multiple recipients in a branch on a 256K link
More than 17x faster transmission for a sender and recipient in the same branch on a T1 link
More than 41x faster transmission for a sender and recipient in the same branch on a 256K link
Up to 97% reduction in bandwidth for these email transmissions

Background

Outlook 2007 and Exchange 2007 have been updated to simplify and integrate communications in the office. Outlook 2007 now has a much richer user interface with plenty of bells and whistles. Another major difference over 2003 addresses the Security and Privacy area. Users now have the ability to encrypt their data between Outlook and the Exchange server. However, in order to enjoy the highest performance with Steelhead appliances, it is recommended that Outlook 2007 users disable encryption (which is enabled by default in Outlook 2007). See the following Knowledge Base article on Riverbed's Technical Support site for information on ways to disable Outlook encryption: https://support.riverbed.com/kb/solution.htm?id=50 1700000008VT8AAM ://https://support.riverbed.com/...

Test Configuration

- Encryption of data between Outlook 2007 and Exchange 2007 was disabled .
- Client OS was Windows XP Professional Client running Office Enterprise 2007.
- Server OS was 64-bit Windows Server 2003 R2 Enterprise Edition running Exchange Server 2007.
- Steelhead 100 appliances running RiOS 5.0.0 #38.
- Simulated WANs = T1 (1536 Kbits/second) and 100 ms latency, 256 Kbits/second and 100 ms latency

Test Methodology

To show the effectiveness of the Steelhead appliances, we performed the same series of tests previously done with Outlook 2003 Cached Exchange Mode testing.

The first test is what we call the " Boomerang " use case where two clients are both in the same branch, one sending a message to the other. The performance metric will consist of the total time required to complete two operations, the Send by client 1 and the Receive by client 2. The Send time is defined by when the Send button is clicked to when the message disappears from the outbox on client 1. The Receive time is defined by when the "updating inbox" message appears and disappears on client 2.

The second test is what we call the " Flood " use case where you have multiple clients in the same branch (three in this case) all logged into Outlook ready to poll for emails. A user on the datacenter side sends a broadcast message that goes to all three of them. We are only concerned with the Receive time for all three clients in the branch since the outgoing broadcast message will not traverse the WAN due to it being on the datacenter side with the Exchange Server. The performance metric will consist of the total time required for all three clients to receive the message. The Receive time is defined by when the "updating inbox" first appears on any of the three clients and when the "updating inbox" message disappears from the last receiving client.

Performance Results - Boomerang

Each test was performed repeatedly over a simulated WAN with 100 ms RTT latency and bandwidths of 256Kb and 1536Kb using a 6.14 MB Powerpoint attachment.

The results show a warm performance improvement of 41x for the 256k case and 17x for the T1 case (previously 33x and 10x for Outlook 2003/Exchange 2003). The bandwidth reduction was in the range of 44-48x (previously 37-43x for Outlook 2003/Exchange 2003).

Performance Results - Flood

Each test was performed repeatedly over a simulated WAN with 100 ms RTT latency and bandwidths of 256Kb and 1536Kb using a 6.14 MB Powerpoint attachment. The results show a warm performance improvement of 53x for the 256k case and 21x for the T1 case (previously 54x and 24x for Outlook 2003/Exchange 2003). The bandwidth reduction was in the range of 36-45x (previously 34-45x for Outlook 2003/Exchange 2003).
Message Edited by bgilbert on 01-28-2009 04:34 PM
‬‪‬‪‬‪