Saturday Jan 28, 2012

Leveraging Infiniband to bypass the BRIDGE

As Deepak mentioned in the previous post, the Tuxedo team has spent a lot of effort in leveraging some of the unique features of the Exalogic platform.  Specifically we've developed support for Remote Direct Memory Access.  This is a feature of Infiniband that allows one node to read or write directly into the memory of another node.  In particular this can all be done from user mode, meaning there is no need to enter the operating system kernel to pass information from one node to another node.

Tuxedo 11gR1PS2 uses this RDMA capability to bypass the BRIDGE process used in a Tuxedo cluster (MP mode domain.)  In standard hardware environments, when a request is made to a server on a remote node, the request is given to the BRIDGE which in turn passes the request to the remote BRIDGE which eventually places the request on the appropriate server's queue.  The reply message takes the reverse path being placed on the local BRIDGE queue, relayed to the remote BRIDGE by a network connection, and then finally placed on the client's reply queue.  In some cases this becomes a bottleneck as the BRIDGE is only partially multi-threaded.  So on high core count systems with a lot of requests being made to remote servers, the BRIDGE creates a throughput bottleneck.  As well the BRIDGE introduces substantial latency as the total round trip requires 4 System V IPC messages and two network messages.  Where a local request/response can be performed in about 35 microseconds, a remote request/response through the BRIDGE can take about 1100 microseconds.  This diagram shows the message flow:

With the BRIDGE bypass feature, a native client uses RDMA to place its request directly on the remote server's queue, and the reply is placed directly on the client's reply queue.  This eliminates two message queue operations and two network operations.  The net result is that throughput increases 7 fold for remote operations by bypassing the BRIDGE and reduces latency from 1100 microseconds to 160 microseconds.  This diagram shows the message flow:

For the next release of Tuxedo sometime this summer, we're planning even more optimizations to achieve even more throughput and lower latency for remote operations.

Tuesday Jan 03, 2012

Tuxedo is now optimized for Exalogic

We are in 2012! Wish you all a very happy new year. In case you missed the announcement, we released a new version of Tuxedo, Tuxedo, middle of December 2011. If you have seen the press release, you know this release includes Tuxedo optimizations for the Exalogic platform. This is first of the blog in a series to talk about Oracle Tuxedo features.

We have changed the guts of Tuxedo to take full advantage of Exalogic technology, specially related to Infiniband. These technologies are SDP (Socket Direct Protocol) and RDMA over Infiniband. All Tuxedo components now can use SDP instead of TCP/IP when communicating with other components running on Exalogic and/or Exadata. Use of SDP generally results in less CPU consumption, leaving more for applications to use. With RDMA over Infiniband, clients/servers on one node can talk directly to servers on the other node without going through the BRIDGE process. This serves two purposes, it eliminates BRIDGE as the bottleneck and reduces the latency of communication by an order of magnitude, in some case to as low as a few microseconds. The benchmarks results from our in-house tests have been shown dramatic improvement in throughput.

In addition, one cool feature Tuxedo release includes is self tuning of SPINCOUNT. Tuxedo now monitors application workload on Exalogic and based on the workload it adjusts the value of SPINCOUNT dynamically for the best use of CPU cycles. Optimized tuning of SPINCOUNT is crucial as SPINCOUNT is used in the context of bulletin board locking and impacts almost each and every operation in Tuxedo. With Tuxedo on Exalogic, there is one less thing to worry about when configuring Tuxedo.

Other features includes in this release are:

- OVM 3.0 certification

- OVAB (Oracle Virtual Assembly Builder) plugin for Tuxedo

- TLOG Store in Database

Keep an eye at this blog for more details to follow!

Deepak Goel

Sr. Director, Tuxedo Development

Wednesday Aug 24, 2011

Oracle Tuxedo at OOW 2011

Oracle Open World 2011 is just around the corner. There will be quite a bit of Oracle Tuxedo specific activities this year. We will host five breakouts, two hands-on-labs and one birds-of-feather session. You will hear what’s new with Tuxedo and see many of the new features in action at the demo grounds. We will kick off the activities with Tuxedo roadmap discussion and delve into new features introduced over the last year, such as deploying existing Tuxedo applications and/or mainframe applications on Exalogic for better performance, reduced complexity and reduced operating expenses. Here is chronological listing of all Tuxedo related sessions:

1. 15730: Oracle Tuxedo: Roadmap and Strategy - Monday, 2pm

2. 15705: High performance web applications with C/C++/PHP/Python - Monday, 5pm

3. 33700: Birds-of-Feather: Road to Oracle Exalogic for Oracle Tuxedo Applications – Monday, 6:30pm

4. 15040: The ART and Practice of Mainframe Migration and Modernization – Tuesday, 4pm

5. 31020: Hands-on-Lab: Mainframe to Oracle Exalogic – Wednesday, 4:45pm

6. 15703: Integrating Oracle WebLogic/SOA with Legacy Mainframe Applications - Thursday, 10:30am

7. 31120: Hands-on-Lab: Develop High performance, service-oriented C/C++ applications for Oracle Exalogic – Thursday, 12pm

8. 15731: Modernizing Mainframe Applications Using Oracle Tuxedo ART 11g – Thursday, 1:30pm

Hope to see you in these sessions and at demo grounds!

Deepak Goel

Sr. Director, Tuxedo Development

Wednesday Apr 20, 2011

Tuxedo is certified on Exalogic and other platforms

[Read More]

Wednesday Oct 06, 2010

TSAM SLA Monitoring

One of the most powerful features of TSAM is the ability to monitor Service Level Agreements. TSAM provides two alerting mechanisms. One allows you define alerts that should be checked by the TSAM console. As data is reported to TSAM Manager (the data collection and storage portion of TSAM) it is compared against alerts that have been defined in the console. This can be a great tool in monitoring your Tuxedo applications.

The one drawback though is that the data has to be reported to TSAM Manager. Given that many Tuxedo applications process 100's or 1000's of requests per second, if the monitoring policies for these kinds of applications enable reporting most of the information that can be collected, it is likely that the TSAM Manager component will be overwhelmed with too much data.

In order to provide finer grained monitoring and ensure that SLAs are being adhered to, TSAM allows alerts to be defined at the plug-in level. The triggering of these alerts is based upon an FML32 boolean expression against the monitoring data evaluating to true. Once an alert is triggered, an alert may be sent to the TSAM Manager, trigger a Tuxedo event to be processed by the Tuxedo EventBroker, and optionally have the message be dropped from the servers IPC queue in the event that there is no point in processing the request.

In my next post, I'll give the details on how these sorts of alerts are defined and some additional ideas about what sort of actions you might want to take place when these alerts are triggered.

Todd Little
Oracle Tuxedo Chief Architect

Tuesday Aug 24, 2010

China and Europe OTN Virtual Developer Day for Tuxedo

[Read More]

Friday Jul 30, 2010

OTN Virtual Developer Day in North America a huge success

If you missed the Virtual Developer Day for Tuxedo sponsored by Oracle Technology Network, then you missed out on a great event. Attendees got to listen to a keynote on how to develop SOA applications without killing your performance. As well they learned what Tuxedo and scripting languages can do for developing agile fast time to market applications using the combination of Tuxedo's robust infrastructure with the quick development time associated with scripting languages. Finally attendees were able to try these things out by executing hands on labs either on the Rackspace Cloud or on their own machines using a downloadable VirtualBox appliance that contained all the necessary software.

The good news is that if you missed the event, you can still go to the event and listen to the sessions and try the hands on labs. Simply visit the event's wiki home page and select Post Conference Content Access.

Friday Jul 23, 2010

Tuxedo at Oracle Open World 2010

[Read More]

Monday Jul 12, 2010

Tuxedo Virtual Developer Days

[Read More]

Tuesday May 18, 2010

Python and Ruby in Tuxedo

[Read More]

This is the Tuxedo product team blog.


« November 2015