At OpenWorld this past month, Oracle® announced a new Exadata X8M system and new Oracle Autonomous Database Cloud.
Oracle Exadata X8M
Oracle re-designed Exadata X8M to take advantage of RoCE for access to NVMe storage and direct RDMA to access Intel Optane DC PM. Exadata X8W ships with RoCE and Optane DC PM for no additional cost and will support up to 1.5TB of DC PM per storage server or 21TB per full rack.
The prior generation, Exadata X8 used InfiniBand for high speed IO access but with 100Gb/sec Ethernet and RoCE (RDMA over Converged Ethernet), InfiniBand is no longer needed. But Oracle has also added Optane DC PM as another tier for data storage that stands between DRAM and NVMe flash SSDs to provide extremely fast access to the hottest data. Optane DC PM is accessed through an entirely different channel which appears to be an extension to the CPU server PCIe bus to the storage servers Optane DC PM, using RDMA.
The new Optane (DC PM) DIMMs are used for both read and persistent write data and form an extension to compute server CPU cache and DRAM. The net result of all this is 2.5X more IOPS (@16M IOPS) than previous generation Exadata and 10X better response time (@19 msec) to database data. Optane DIMMs are also used to increase log commits. Exadata X8M uses RDMA and writes to multiple (2+) Optane DIMMs on different storage servers to accelerate data base commits resulting in 8X faster log writes. All this added performance is available to all customers without the need for software change as it’s transparent to customer access. Oracle showed some comparisons of Exadata X8M performance for an 18B row database query. The query took 130 seconds on previous gen Exadata X8 and only took 11 seconds with Exadata X8M. They did the same query with Exadata X8M using a preview version of Oracle 20C which supports deep Intel vectorization integration which only took 3 seconds.
Oracle Autonomous Database in Cloud
Oracle has implemented a new autonomous cloud database, with two distinct versions, one targeted for data analytics/warehouses, the Autonomous Data Warehouse (ADW) and the other targeted for OLTP, the Autonomous Transaction Processing (ATP). Both provide a fully automated database lifecycle management for a self-driving/-securing/-repairing/-tuning database solution. Oracle has been on a long journey to automate management and optimization of their database software and hardware, ATP and ADW are the culmination of all this work.
ATP and ADW both minimize all human interaction to use enterprise class databases. Both ATP and ADW provision databases rapidly by creating an Exadata cloud infrastructure RAC database with Active Data Guard standby, automatically secure by monitoring for threats, applying security updates and stopping data exposures with DB Vault and data encryption and autonomously manage by performing all OS and SYSDBA operations, tuning, and updating all software, automatically recover from site outages, autonomously scale databases using serverless infrastructure and continuously optimize memory, data formats, indexes, parallelism and plans for each workload. ADW and ATP are specially designed to optimize for their respective workloads .
Oracle presented some statistics comparing ADW and ATP to manually managed/optimized databases and showed that ADW and ATP consistently resulted in lower plan execution times for a number of verticals and workloads.ATP and ADW are also offered with Dedicated Exadata Cloud Infrastructure which provides guaranteed 99.995% availability and fully isolated compute, storage and cluster interconnect.
Exadata X8M is an interesting approach to adopting Optane DC PM and only available to those vendors that own the complete hardware and software stack. The increased IOPS and drastically reduced latencies are proof positive of what total stack owners can do when innovating with the latest technology offerings. Optane DC PM was only released in April of this year.
ATP and ADW are also interesting as the high point of a journey that every vendor is on today. There’s a lot of AI embedded in ATP and ADW to make all of this work autonomously. And yet it’s another good example of what any corporation can do when they take advantage of the newest AI-ML technologies and have lots of data to train with.
[This storage announcement dispatch was originally sent out to our newsletter subscribers in September of 2019. If you would like to receive this information via email please consider signing up for our free monthly newsletter (see subscription request, above right) and we will send our current issue along with download instructions for this and other reports. Dispatches are posted to our website at least a quarter or more after they are sent to our subscribers. ]
Silverton Consulting, Inc., is a U.S.-based Storage, Strategy & Systems consulting firm offering products and services to the data storage community