Here we are at the end of another ECO conference. Yesterday I presented "Putting on the Flash" describing my testing of the flash cache options in 11gR2. Today I present an updated "Exadata: A Critical Review" which now includes some data on the X3.
It has been perfect weather here but they are expecting thunderstorms today, hope my flight home isn't delayed. The IBM conversion of TMS is going well so far from what I can see from Atlanta. The usual issues with organization and understanding how the we verses they way of doing things are happening, but I don't believe it is anything that can't be resolved.
The biggest new feature in Oracle12c seems to be the concept of pluggable databases. I will review this feature once I can actually play with it, but since I haven't been invited to the beta party I will have to wait like the rest of you, unless IBM can use their partner clout to get an advanced copy. I am trying to get the lab in Houston to provide me with a RamSan820 and an IBM server so I can begin testing for performance and other topics with our new technologies. Keep watching for anything interesting I find!
Thursday, October 18, 2012
Tuesday, October 2, 2012
X3 the Next Generation
So Larry has announced the next big thing, the X3 series of the Exadata. After reviewing the specifications I see mostly it is providing 4 times more flash cache, an improved caching algorithm and the capability to buy a 1/8th rack configuration.
When will they finally ditch the disks and go all flash? Face it, we don't need more flash cache if the main stoarge is flash.
The graph above demonstrates what I talking about. With a 40 gig SGA I set the flash cache to 90 gig. As you can see using flash (a RamSan810) as the main storage and running a standard TPC-C benchmark the performance difference is negligable with the zero level of flash cache actually performing better in the upper range. Why pay for 12 disk licenses in perpetuity when you don't need to? If you get the new Exadata storage cell with the increased flash cache, that is what you will be doing.
Larry also placed great store in the HCC compression technology. However you must realize HCC requires you do a direct load of the data and is virtually useless in OLTP, in fact, Oracle themselves say not to use it in OLTP but to instead use standard advanced compression.
For what the Exadata and and now the X3 is designed for, data warehousing and environments with highly redundent, stable data, it will be a great performance booster, but for the rank and file OLTP databases I see it providing marginal performance increases at best.
When will they finally ditch the disks and go all flash? Face it, we don't need more flash cache if the main stoarge is flash.
The graph above demonstrates what I talking about. With a 40 gig SGA I set the flash cache to 90 gig. As you can see using flash (a RamSan810) as the main storage and running a standard TPC-C benchmark the performance difference is negligable with the zero level of flash cache actually performing better in the upper range. Why pay for 12 disk licenses in perpetuity when you don't need to? If you get the new Exadata storage cell with the increased flash cache, that is what you will be doing.
Larry also placed great store in the HCC compression technology. However you must realize HCC requires you do a direct load of the data and is virtually useless in OLTP, in fact, Oracle themselves say not to use it in OLTP but to instead use standard advanced compression.
For what the Exadata and and now the X3 is designed for, data warehousing and environments with highly redundent, stable data, it will be a great performance booster, but for the rank and file OLTP databases I see it providing marginal performance increases at best.
Monday, October 1, 2012
Day One of Open World
I arrived in San Francisco at about 11 am local time on Sunday September 29, grabbed my luggage and jumped in a cab to make to my first co-presentation, "VM Experts Panel" at 12:30 at Moscone West. Immediately following the VM Experts panel, was my second presentation, " Database Performance Tuning: Get the Best out of Oracle Enterprise Manager 12c Cloud Control" with Tariq Farooq and Rich Niemiec. Both presentations were well attended with standing room only for the second.
After finishing the second presentation I checked into my hotl, but only had a half hour before I had to get back to the Moscone center for the opening day Keynote. Fujitsu's Corporate SVP Noriyuki Toyuki opened the Keynote with Fujitsu's vision of the future heavily pushing cloud initiatives with their smart cities and agricultural and medical applications.
The final half of the Keynote was all Oracle CEO Larry Ellison with Ellison making the first announcement of the IaaS (infrastructure as as service) initiative from Oracle were they will provide you with equipment either hosted or on your site and only charge you for what CPU and other resources you actually use. Larry also hinted that there will be new changes to the Exadata line of products "eliminating IO latency" look for something involving SUN SSD PCIe cards or form factor SSDs in the Exadata storage cells would be my guess.
Following the keynote I had to run to catch the bus, which I missed, and ended up taking a taxi, to the ACE dinner at the St Francis Yacht Club and finished the evening with a rousing political debate with my friend Steven Feuerstein. Needless to say I ended up going to sleep at 10 pm local time after an exhausting but rewarding day.
After finishing the second presentation I checked into my hotl, but only had a half hour before I had to get back to the Moscone center for the opening day Keynote. Fujitsu's Corporate SVP Noriyuki Toyuki opened the Keynote with Fujitsu's vision of the future heavily pushing cloud initiatives with their smart cities and agricultural and medical applications.
The final half of the Keynote was all Oracle CEO Larry Ellison with Ellison making the first announcement of the IaaS (infrastructure as as service) initiative from Oracle were they will provide you with equipment either hosted or on your site and only charge you for what CPU and other resources you actually use. Larry also hinted that there will be new changes to the Exadata line of products "eliminating IO latency" look for something involving SUN SSD PCIe cards or form factor SSDs in the Exadata storage cells would be my guess.
Following the keynote I had to run to catch the bus, which I missed, and ended up taking a taxi, to the ACE dinner at the St Francis Yacht Club and finished the evening with a rousing political debate with my friend Steven Feuerstein. Needless to say I ended up going to sleep at 10 pm local time after an exhausting but rewarding day.
Subscribe to:
Posts (Atom)
