that writes backup and archive copies of data directory to any object storage target — whether it is public cloud or on-prem
[Webinar] Protect Your Most Trusted Mainframe Data Across the World in an Age of Heightened Cyber-security
that writes backup and archive copies of data directory to any object storage target — whether it is public cloud or on-prem
A US-based bank was using the Oracle StorageTek Virtual Storage Manager (VSM) System to write to on-prem tape storage and wanted to reduce the amount of time it took to complete its backup jobs.
Model9’s Cloud Data Manager for Mainframe replaced the bank’s VTL and backup software and instead began writing data directly from the mainframe to object storage in the Azure public cloud.
What is the technological environment?
• z15 with 2 x zIIP engines
• z/OS Version: V2R4
• Virtual Tape: StorageTek VSM6
○ 4 X 8Gb FICON ports
• MF Network: 2 x 1Gb OSA ports
• Azure Network: 1 x 1Gb
What is being compared?
IBM – Oracle Solution | Model9 – Azure Solution | |
---|---|---|
Target Storage: | Oracle StorageTek Virtual Storage Manager System (VSM) | Azure Blob Storage |
Backup Management SW: | Multiple DFDSS backup JOBs | Model9 Backup Policy |
By using the Azure Cloud as a target instead of on-prem tape storage, Model9 was able to push data over TCP/IP and write directly to the Cloud. That enabled enough parallelism to use all of the bank’s available bandwidth, maximizing throughput performance.
Instead of taking an entire weekend day to complete writing a backup job to VTL,
Model9 was able to write to Azure Cloud in just 1h 11min.
In fact, the bank’s ingestion throughput capabilities increased so dramatically that it decided to increase its Cloud network throughput capabilities to take full advantage of what was now possible with Model9.
Performance KPIs:
Action | IBM – Oracle Solution | Model9 – Azure Solution |
---|---|---|
Full volume dump backup of 2.2TB |
Elapsed time: 24 hours |
Elapsed time: 1h 11min |
Ingestion throughput: 27MB/s |
Ingestion throughput: 547.96 MB/s |
|
FICON throughput: 27MB/s |
Network throughput: 102.89 MB/s |
|
Data sets backup of 18.46GB |
Elapsed time: 5 min. 34 sec |
Elapsed time: 36 sec |
Ingestion throughput: 56MB/s |
Ingestion throughput: 456MB/s |
|
FICON throughput: 56MB/s |
Network throughput: 23.7MB/s |
|
DB2 recovery of 4.5GB |
Elapsed time: 02 min 13 sec |
Elapsed time: 01 min 37 sec |
Restore throughput: 34MB/s |
Restore throughput: 47MB/s |
Data sets backup
DB2 Image Copy Archive
An EMEA bank was looking to replace its IBM virtual tape libraries, and its data management software: DFSMShsm, DFSMSrmm, and IBM Security Key Lifecycle Manager (ISKLM). It wanted to increase throughput and modernize its mainframe infrastructure to write directly to IBM Cloud Object Storage (COS).
To achieve this, it decided to test the performance of the Model9 Cloud Data Manager – IBM COS solution vs. the IBM TCT – COS solution.
What is the technological environment?
• z15 with 4 x zIIP engines
• z/OS Version: V2R4
• Network: 2x 10Gb OSA ports
What is being compared?
IBM Solution | Model9-IBM COS Solution | |
---|---|---|
Target Storage: | IBM COS | IBM COS |
Archive Management SW | IBM DFSMShsm | Model9 Cloud Data Manager for Mainframe |
Additional Hardware Features | IBM DS8900 and Transparent Cloud Tiering (TCT) | None |
Because the bank wanted to go to a solution with no physical tapes, IBM TCT required additional DASD space to accommodate all writes and reads including DFSMShsm. In contrast, Model9 does not require DASD space for backups.
The bank tested IBM TCT with IBM COS on-premises as the target storage for TCT, and used a large DASD pool for DB2 Image
Copies and let DFSMShsm archive the data to IBM COS using TCT. It found that Model9 offered better performance.
It’s also worth noting that when writing backups from VTL/Tape, TCT does not replace DFSMShsm or DFSMSrmm.
In contrast, Model9 is a single software solution with no additional hardware footprint — making it a much cheaper and easier solution to implement.
Performance KPIs:
Action | IBM Solution | Model9 – IBM COS Solution |
---|---|---|
Image Copy Archive |
Size: 1.5TB |
Size: 1.1TB |
Elapsed Time: 6 hours |
Elapsed Time: 21 min |
|
Ingestion throughput: 70MB/s |
Ingestion throughput: 908MB/s |
Model9 reached its peak while archiving 1.1TB in 21 minutes, with a throughput of 908MB/s:
Cloud Storage | IBM COS |
# IBM COS Nodes | 2 |
# Model9 Agents | 2 |
Elapsed time (sec) | 1270 |
Total Data GB | 1126.4 |
Size In Cloud GB | 469 |
Ingestion Throughput MB/s | 908 |
Network Throughput After Compression MB/s | 378 |
Full volume dump backup of 2.2TB
Model9 Cloud Data Manager for Mainframe maximizes your available bandwidth, increasing data throughput speeds
when making backup and archived copies of data. It accomplishes this by:
- Achieving parallelism by using multiple TCP/IP streams to
send data in multiple chunks to any object storage target,
instead of writing to tape serially. - Compressing data using zIIP or zEDC capability when
available before pushing it out of the mainframe, minimizing the amount of data being sent and increasing the effective capacity of your network connection. For example, compressing data at 3:1 allows a 1Gb line to now handle 3Gb.
- Using zIIP engines instead of mainframe MIPS to do
the work, reducing CPU overhead that can be used for
other mainframe processes. - Spreading the workload across the entire mainframe
environment using a multi-agent load-balancing policy. - Writing multiple data sets, and breaking large data sets
up into chunks.