Registration was simple but due to long lines it took about 30 minutes. 20 minutes to get the badge in the level 1 Moscone west and 10 minutes in the level 3 for the bag and alumni jacket.
The first session was scheduled for 12:30pm ..we had a good 2 hours..utilized it to the fullest extent by roaming around the market street and taking an early lunch in a Mexican restaurant.
I managed to attend several sessions today.. Here is the snapshots of the sessions that I attended today...
Session 1: Field report: Extended distance RAC 11g on sparc technology
12:30p to 1pm
Speakers: Andreas Ellerhoff, Deutche Messe and Mike Dietrich, Oracle Corp
Interestingly the first session that I attended was in room 2010! Agenda sounded great but the session ended up a little disappointment as the session and slides contained only the technical specifications and not more details.
Extended distance RAC, the speaker explained was of three data centers each 0.7 miles apart with 2 nodes (node 1 and 3) in the data center 1 , 1 node (node 2) in the data center 2 and the centric storage backup server in the data center 3.
The 8 reasons as per the speaker that triggered 11g upgrade are:
~ Harmonization
~ Security
~ Reorg (reduced the emc disk storage by half)
~ technological improvements
~ grid control for easy monitoring
~ significant performance improvements
~ use of further new feature
~ low cost for the upgrade (?!?)
No time was given for Q & A as the session was allotted only 30 minutes. Kudos to the speaker for accepting to do this session in 30 minutes!
Session 2: Centralized HA database on 8 node RAC cluster using 11gr2
1pm to 1:30pm
Speaker: Christian Trieb, Paragon data systems
The speaker laid out high availability, good performance and short time frame for migration as goals for their upgrade.
Their migration plan sounded very simple based on the high level steps that speaker shared in the session...
~ export .. 2 hrs
~ import ...with parallel 32, data plus index and no stats
~ gather stats .. 3 hours
Speaker was honest in saying that they were able to perform migration only on their 5th attempt.. 1st and 2nd attempt got cancelled due to lack of time for app testing, 3rd attempt was cancelled due to delay in getting bug fix for an oracle internal error (kjmsendscq:timeout) and 4th attempt was cancelled due to problems with backup.
Btw three bugs were referenced for their ORA-00600 internal errors. I jotted down the bug numbers but not sure whether I got it right or not giving it here any way.. 9831998, 9696187 and 987053
Here is the comparison of their pre-mig and post migration configuration..
6 nodes rac+dg versus 8 nodes rac+dg
10203 versus 112010 plus 9778840 patch for grid infra home
32 gb ram versus 96 gb ram
12g sga versus 20g sga
Linux 2.6.9(4 nodes) and 2.6.18(2 nodes) versus 2.6.18(8 nodes)
Issue with netapp backup snap mgr post migration.. As they are seeing performance issue if integrated with rman. With rman they are getting done in 40 minutes and without rman integration they are getting done in 10 minutes.
Similar to session1, no time was there for Q & A as the session was scheduled only for 30 minutes
Session 3: IOUG: DBA Fusion --- an introduction to oracle enterprise manager 11g
1:30p to 2:30p
Speaker: Ray Smith, Portland General Electric
At the start of the session itself the speaker set the expectation that this session is about oracle management server installation and configuration and not about the OEM console or database.
It was interesting to know from the speaker that there are about 60 sessions pertaining to OEM in this year's open world.
The high level steps for the install were explained as listed below:
Step 1: gather documentation ...( he also recommended oracle-wiki.net for a step by step documentation on install) middleware home, inventory home and grid infrastructure directly under oracle base folder.
OMS home, agent home and config file under middleware home.
Step 2: Install and patch fusion middleware (tip suggested was to use the jdk that get shipped from the product instead of the existing jdk.
Step 3: create repository db (compatibility as per note 412431.1)
Step 4: install oms
Other interesting info from the speaker to share are.. Setup communication between your oms server with oracle support and put oms and repository db on separate servers due to excessive java memory usage by middleware home
Unlike the session 1 and 2 above, time was given for Q & A.
Session 4: Resolving free buffer waits event
3p-4p
Speaker: Craig A Shallahamer, orapub
Craig is doing 4 sessions this year, two of them regular sessions and two of them unconference sessions.
He also utilized this session for a little marketing of his books .. 1: oracle performance firefighting and 2: forecasting oracle performance.
There was a raffle for two books in this session.
Craig's explanation of the buffer cache, buffer cache header, cache buffer chains, lru, lru-write was excellent. His style of lecture was impressive. The sad part was this session was forced to stop 5 minutes before the schedule to allow the next session to start on time and was warned 12 minutes before the end of the session. There was no time reserved for Q & A but during the lecture itself, Craig encouraged questions from the audience.
After explaining the reasons for free buffer wait with great detail and animation, He summed up the solution as follows:
Increase buffer cache
Tune SQL to reduce physical I/0
Review application design/logic to find the DML to check it's work load balancing issue
Analyze SQL statements with hi consistent reads
Decrease _db_large_dirty_queue
Increase _db_writer_max_time ( not sure I got this right or not!)
Session 5: Oracle RAC expert panel
4p-5p
Panel: serge del rio (moderator), Markus (Product mgr), Sri Subramaniam, Nitin vengulekar (Product strategy), Phil Newman and Krishnadev Telikicherla
This is a great session as it was purely based on questions and answers format. The scheduler builder had this session as customer panel instead of experts panel. At the start a request was made to keep the questions unrelated to any SRs opened and/or bug specific.
Oracle RAC SIG election results were announced prior to the session. Alex gorbachev is the new president of the RAC SIG.
Questions started very dry and picked up momentum in the middle.
Application readiness document/guidelines to identify how application will perform in rac or nonrac... Panel's answer was.. if it scales well as single instance, it will scale well as rac. The main concern mentioned was regarding the sequences and the index on sequence.
Regarding my question of rac-aware connectivity details for jdbc thin drivers, the behavior is same in 11.2 as 10.2 and a white paper on universal connection pooling was suggested for reading.
Regarding the question on parallel queries within the instance and across the instances... The answer given was it depends and test it before you use it!
Recommendations given for the number of nodes in a rac environment.. Minimum is more than 2 (i.e three nodes) and max is 100 nodes. But 100 nodes in reality is a complex thing if your app is scaling upto 100 nodes, chances are it may need more and you can. Not go beyond 100 nodes.
Panel suggested 36 nodes as a realistic upper limit?
Here is the most interesting information that was told by the panel...
11.2.0.2 changes:
Lot of Bug fixes
Quality of service management new feature to be announced in this open world.
ASM enhancements.. Power limit upper limit increase from 11 to 64, Asynchronous rebalance and Enhanced cp command.
Here is my favorite... Clusertware will not reboot the node for interconnect issues or any other os issues, instead it will only stop the clusterware stack.
This is based on customer demand as several customers have requested availability for non cluster resources running on the same node.
Side bar: we were in the same shoes, we had app servers running on local zones and global zone dedicated for oracle rac.. But rac reboots caused outages to our app servers running on the same hardware in the local zones. We migrated the app servers to a different hardware and removed all the local zones from the rac nodes leaving just the global zone for cluster ware and it's resources.
Key note session:
Speakers: HP ( Ann Livermore and Dave Donatelli) and Larry Ellison
Hp's agenda was focused on datacenter solutions for today and tomorrow: App modernization and converged infrastructure.
Other facts that was interesting to note : this year's open world is the biggest in history in terms of attendance .... 41 thousand attendees.
Larry started his speech by defining the cloud computing by differentiating the amazon ec2 and salesforce.com saas.
Major announcement that he made was the availability of exalogic elastic cloud ...cloud in a box.
Quote of the day: spend less if you want better performance.
Exalogic machine is 4 times lower cost compared to ibm's powerful power 795. The price quoted in the slide.. $107500 for exalogic machine and 4440000 for power 795
Exalogic machine key features summary..
Fastest java performance
Elastic capacity on demand
Fault tolerant
Scalable
Secure
Easily Maintained.. One patch / one file for all software components in the machine
No single point of failure.. Redundancy all layers
Standard configuration of all customers..helps oracle deliver and test better..discover the issues faster and fix bugs
Internet application - 1 million http requests per second
Messaging application - 1.8 million Messages per second
30 computing servers
360 cores
2.8 tb dram
460 gb solid state disk
1.2 microseconds latency
4 tb read cache
Clustered for ha
40gb infiniband, Eliminate buffer copies, use large pkt size
Transaction affinity to db node
New tagline .. Hardware and software engineered to work together
The second announcement made was the announcement of oracle unbreakable enterprise kernel ... Much faster than red hat, gets community and oracle enhancements much faster. Oracle remains committed to red had compatibility
The availability of the new exadata top of the line oltp database machine will be announced later this week.
The final portion of larry's keynote was the preview of fusion app 5 year development and a demo by Steve Miranda.
- Posted using BlogPress from my iPad
Location:San francisco, CA
.jpg)
