Having ultra-fast performance for your databases is critical – especially if you’re managing all your customers’ most mission-critical data in your database. But when it comes to offering a cloud-based solution for customers, managing this data and achieving the level of performance needed becomes even more complicated.
This is what one product lifecycle management (PLM) company discovered when it adopted a cloud-based solution on the Microsoft Azure cloud. Cloud vendors put throttles on their maximum speeds making it impossible to get lightning-fast performance… unless you’re willing to pay out of the ears for it.
The company decided to try out the Silk Platform with its Oracle workloads. Silk sits between the cloud infrastructure and workloads. On average, Silk offers customers 10x faster performance than they can achieve with native cloud alone.
A 1TB Oracle pluggable database (PDB) was provisioned for these tests. And the results were staggering.
Test #1: Provision New PDB
The first test that was run was for the provisioning of a new PDB, required as part of the process of onboarding new customers. The process involved the creation of a new PDB of approximately 1TB in size.
Without Silk, this onboarding process took 59.9 minutes. But with Silk, creation time was only 10.58 minutes. That’s an 82.3% decrease in the time it takes to provision a new 1TB PDB.
Test #2: Data Pump Import (Using Parallelism)
The next test was for a Data Pump Import, a process used to seed new customer PDBs with initial data.
Without Silk (with Oracle redo logs located on a P40 disk), the import time was 1:20:46 hours. But with Silk, the import time was only 27:17 minutes. That’s a 66.2% decrease in the time it takes to run the Data Pump Import.
Test #3: Data Pump Export
The third test was for a Data Pump Export, which is a process used for database logical back-up purposes. The source was an 1TB sample database, with parallelism set to a degree of 4.
Without Silk, back-ups were written to /backup LVM using two P40 Azure disks with the source database being one P40 Azure disk. The time required to complete the export was 23:33 mins.
With Silk, the backups were written to /backup on a Silk volume with the source database being a Silk volume as well. The time required to complete the export was 8:42 mins
Meaning with Silk there was a 63% decrease in the time it takes to run Data Pump Export for a logical back-up!
Test #4: Full Physical Backup
But what about a physical back-up, you ask. For the next test, the company ran a PDB RMAN FULL backup. The dataset was a sample database of approximately 1TB in size. Again, the parallelism was set to a degree of 4.
Without Silk, the backups were written to /backup LVM using 2 P40 Azure disks, with the source database on one P40 Azure disk. The process took 36:14 mins.
With Silk, backups were written to /backup on a Silk volume with the source database also being a Silk volume. The backup only required 11:52 mins
With Silk, the company is able to decrease the time it takes to do a physical backup by 66.8%.
Test #5: PDB Local Clone
The fifth test was for a parallelized local database cloning operation. The local PDB copies were within the same CDB with PARALLEL=8. The data came from a consistent source of a freshly Data Pump imported PDB with a size of approximately 1TB. A test like this is useful in situations where a full PDB online backup is needed prior to an application upgrade.
Without Silk, the local clone took 01:17:04.16 hours.
With Silk the results were dramatically faster at 00:10:37.75 mins
That’s an 86% decrease in the time it takes to clone a local PDB!
Test #6: PDB Remote Clone
The final test was for cloning the Oracle pluggable database over the network to a remote host. The data was copied through DB_LINK with PARALLEL=8. The source server and target server were based in the same Azure availability zone.
Without Silk, the process took 01:17:09.03 hours
With Silk, the process took only 00:16:53.45 mins
That’s a 77% decrease in the time it takes to create the remote PDB clone!
Ready to start seeing the same ultra-fast performance? Click here to learn more.