Join Transform 2021 for the most vital themes in enterprise AI & Data. Learn more.
This short article is portion of the Technology Insight series, created feasible with funding from Intel.
For the final two years, we’ve been writing about Intel Optane persistent memory, being aware of that scores of evaluations and pilot projects are underway in enterprises about the globe. But quotable case research have been uncommon. That changed lately when Intel publicly presented examples of Optane’s actual-globe suitability and worth.
During a higher-profile occasion, Boeing revealed its production information center deployment of Intel Optane persistent memory by means of Oracle’s Exadata X8M database servers. Not lengthy just after, VentureBeat snagged an exclusive meeting with Maruti Sharma, Boeing’s chief architect for digital widespread services and an associate technical fellow inside the aviation giant. Sharma manages Boeing’s enterprise databases and other information management services. He’s the excellent particular person to take us inside this project and reveal the hands-on specifics of an enterprise-scale Optane deployment.
Key Points
- Boeing’s thousands of Oracle databases have been foundering on outdated commodity servers, though information demands kept rising.
- The enterprise decided to consolidate almost one hundred database and storage servers onto only 3 racks of Oracle Exadata X8M systems supporting Intel Optane persistent memory.
- The move helped drive overall performance improvements of 2x to 10x in Boeing’s database operations.
Boeing’s need to have
Boeing manufactures on each U.S. coasts, which aids drive the company’s need to have for cloud-based information management. All told, Boeing runs more than 21,000 databases, roughly 5,000 of which are Oracle. Given how tempting it is to deploy databases with devoted server stacks, it is simple to consider sprawl and redundancy accumulating more than time. So, as in lots of enterprises, decreasing the information center server footprint is one of Sharma’s prime objectives.
Not surprisingly, provided the enormous quantity of databases, Boeing had an ocean of facts it could barely sip. An instance:
“Every commercial flight is spitting out tons of data, most of which Boeing hadn’t even started processing,” Sharma explains. “We need a lot of infrastructure to process that data, so we can mine useful information to grow the business. So if a part is going bad, the plane can relay that information to the airport where the flight is going to land. Engineers and mechanics would be ready with that part. So time-to-service is lower. The mechanics spend less time maintaining that plane. And the airline wants the plane to be in the air as much as possible. Better data management means better value for our customers.”
Whether addressing information from the air, factory, or provide chain, Boeing saw a increasing need to have to deal with database processing in actual time. The company’s workloads generally resembled on the web transaction processing (OLTP) datasets, equivalent to these applied by monetary institutions and choice help method (DSS) datasets, which are ordinarily study-only. Boeing generally handled each forms concurrently in mixed-workload scenarios, particularly when combining with other datasets coming in from provide chain partners.
One of Boeing’s database projects involved an operational information shop (ODS) operating Oracle actual application clusters (RACs) on commodity systems. Including dev and test systems and the production atmosphere, the ODS infrastructure consumed roughly one hundred servers. Each cluster server ran OLTP and DSS workloads with 1TB of memory and more than 1PB of aggregate storage. As Sharma describes it, cluster overall performance was “not at par”, and the hardware had reached its finish of life.
New tech for new possibilities
Boeing sent out quite a few RFPs and in the end examined 3 alternatives: an HPE answer, commodity hardware based on the newest-generation Intel processors, and Oracle’s Exadata Database Machine X8M. One of the latter’s major benefits, Sharma says, was its incorporation of 1.5TB of Optane persistent memory for each and every server. Optane’s DRAM-class latency and greater-than-DRAM capacity points created it an excellent match for Oracle’s OLTP-form workloads, which advantage from incredibly speedy access to little chunks of information from cached storage. (In the X8M’s architecture, persistent memory occupies the initially of 3 automatically managed storage tiers, followed by NVMe-attached NAND and lastly difficult disk storage.)
Boeing’s industrial operations gathers all the company’s manufacturing and provide chain information. It’s then aggregated into a information shop, exactly where factory floor transactions are integrated with provide chain information. Boeing makes use of the benefits of these integrations to make informed choices on what inventory is essential for each and every distinct plane.
“We need to get data from storage faster and keep it closer to compute,” explains Sharma. “How much needed data can reside in memory? When data is closer to compute, how fast can I process it? With the introduction of persistent memory and 100 gigabit-per-second RDMA over converged Ethernet (RoCE) network fabric — which lets nodes request data directly from PMem rather than going through the entire stack — we saw an opportunity to eliminate most of the latency.”
To be clear, Optane “PMem” does not replace DRAM. Rather, it serves as higher-capacity volatile method memory though bumping DRAM into a higher-speed caching function. Or it can act as a super-speedy, non-volatile (persistent) cache for storage. Boeing primarily employs the latter, by means of Oracle’s information accelerator functionality.
However, Boeing has a second use in “persisting” (saving into non-volatile storage) Oracle redo logs — a essential step ahead of a transaction can be committed. Redo logs commonly get persisted to standard SAN storage, which introduces substantial latency. That step previously accounted for a lot of Boeing’s lag, particularly considering the fact that the group’s redo lot sizes typical about 24GB. Trying to persist that quantity of information regularly adds considerable method delays.
Thanks to Intel’s App Direct Mode for Optane persistent memory, and Oracle’s current help for App Direct in its platform, Boeing could address each volatile and persistent models simultaneously, says Sharma.
From deployment to benefits
After months of in depth testing, Boeing deployed its new Exadata servers into production in June, 2020. Teams consolidated almost one hundred commodity servers down to only two Exadata racks with eight servers each and every, and a pair of half racks divided across two information centers.
“Overall, based on how much value we could get from any of the options, the Exadata with persistent memory stood out most,” says Sharma. “It was integrated with other Oracle internals, like Oracle Linux and GoldenGate, that we use heavily to bring in data from our OLTP environment.”
According to Sharma, Boeing encountered no concerns through deployment or even a need to have to adapt its computer software to accommodate persistent memory, as Oracle had currently performed this work with Intel. The only additional labor arose from Boeing’s policy against third-party racks into its information centers. As a outcome, Oracle had to re-rack its Exadata systems into Boeing’s personal racks more than the course of quite a few days.
In the months considering the fact that initial deployment, Boeing reports 2x to 10x productivity gains by switching to the Exadata X8M platform. The most significant database operations improvements more than the previous commodity infrastructure came from bolstered redo log overall performance, Sharma says, adding that other examples abound.
“When we run our batch processing, multiple jobs run overnight. Workloads that consistently used to take 14 hours now take about two hours. This really matters because of various work shifts coordinating across time zones. It becomes very challenging when a shift starts and needs results from another group that hasn’t finished. With jobs finishing faster, our shifts can make better decisions.”
Going forward: Doing more with fewer sources
Despite its considerable server consolidation, Boeing says it nevertheless has ample capacity left in its new Exadata answer. This opens the door to taking on more workloads from other tasks or groups. Sharma expects container technologies to play a function in additional consolidation, permitting engineers to cleanly separate, say, manufacturing workloads from the provide chain, engineering, or analytics. Containers could also assistance with compliance in Boeing’s government operations, he adds.
Beyond consolidation and information isolation, the enterprise says it can now retain and handle workloads with fewer sources. For instance, rather than needing separate administrators to handle unique layers of the answer (storage networking, and so on.), one consolidated group now can handle the complete stack. Sharma says this becomes doubly vital due to the fact Boeing’s Oracle databases run on a mix of Linux, IBM AIX, HP-UX, and other operating systems. Having one typical platform reduces spending on sources and infrastructure. Again, it is about the efficiencies of consolidation.
“There has always been a race between the different components of the infrastructure stack,” Sharma notes. “These advances in compute and persistent memory allow customers like us to process more data in a timely fashion. Data is exploding, and so is the demand for storing, retrieving, and processing the data set. These innovations will give us more leverage in consolidating workloads and doing more data analytics locally. Especially with container technology onboard, we can bring in petabytes of data for processing in one location and help drive the business.”