Around the Storage Block
cancel
Showing results for 
Search instead for 
Did you mean: 

The Relevance of Data Integrity in the All-Flash Data Center

StorageExpert

 

End-to-end data integrity with HPE 3PAR StoreServ flash storage.

More than 2.5 exabytes (a billion gigabytes) of data are created every day. The U.S. government claims to have a data center capable of storing a Yottabyte (a unit of information equal to one septillion (1024) or, strictly, 280 bytes).

3PAR_Flash_blog.jpg

Yes, these types of numbers are truly mind–boggling. There seems to be no end to the amount of data we humans store. And with an evolving socially connected world, data becomes all that more pervasive. This all begs the question: what is done to ensure data stored and then read back is exactly what it supposed to be?

This is an important question to answer as the transition to all-flash datacenter brings consolidation where a single array is now handling hundreds of thousands of IOPS, and with technologies like deduplication sharing application data, the pressure on a single system to guarantee consistency is immense so eliminating any form of data inconsistency is paramount.

HPE 3PAR answer: Persistent Checksum

One of the cool features of the HPE 3PAR StoreServ hardware architecture is called HPE 3PAR Persistent Checksum. It’s based on the T10 committee (the committee for SCSI Storage Interfaces) of the International Committee for Information Technology Standards proposed end-to-end data integrity protection – from host CPU to media. Its job is to protect the integrity of data from media or transmission errors. HPE 3PAR StoreServ 8000 and 20000 support this technology from the host HBA, thru the controllers and to the backend media; HPE 3PAR StoreServ 7000 and 10000 support this feature from the array controllers to the backend media.

To put HPE Persistent Checksum to the test, we at HPE commissioned Demartek to demonstrate the effectiveness of this solution at detecting and preventing silent data inconsistency. Demartek examined the function and performance of an all-flash HPE 3PAR StoreServ 8450 Storage array in an HPE StoreFabric Gen 5 SAN while intentionally introducing data transmission errors into the fabric. The results from the independent testing are fantastic news for 3PAR users.

Here’s a look at the key findings from the Demartek report. You can also read the complete report: Demartek Evaluation of HPE 3PAR Persistent Checksum

Key findings of the Demartek evaluation of HPE Persistent Checksum report

  • Data integrity handling must change in response to technology and performance improvements in enterprise storage.
  • HPE 3PAR Persistent Checksum effectively identifies and corrects data integrity issues in the SAN, preventing potential application impacts.
  • HPE 3PAR Persistent Checksum logic had virtually no impact on the all-flash performance of the HPE 3PAR StoreServ 8450 Storage array.

Putting 3PAR Persistent Checksum to the test

The test configuration consisted of a HPE 3PAR StoreServ 8450 All-Flash Array, a Viavi Solutions Inc. Xgig® 1000 analyzer/jammer, HPE StoreFabric SN6000B 16Gb Fibre Chanel (FC) switch and StoreFabric FC HBAs from different OEM vendors. (See Figure 1.) 

Demartek tested 3PAR Persistent Checksum by injecting bit errors into A real FC data stream using “a Viavi Solutions Inc. Xgig® 1000 analyzer/jammer and intentionally silently changing the bits in bi-directional data frames so that SCSI CRC checks would not detect the changes.”

The application environment leveraged a Windows® Server virtual machine supporting an OLTP database application in a VMware ESXi environment. Errors were injected with Persistent Checksum enabled and disabled. With Persistent Checksum disabled the injected errors were so bad that the database and virtual machine (VM) crashed in about 2.5 minutes, and with Persistent Checksum enabled the database and VM application continued to run unabated.

3PAR_Demartek_f.jpg

Figure 1:  Evaluation environment topology

What is the impact of 3PAR Persistent Checksum on storage system performance?

In addition to testing Persistent Checksum against errors responsible for silent data corruption, we wanted to ask the obvious question, how does Persistent Checksum affect performance in terms of bandwidth and response time?  Demartek confirmed that by enabling and disabling 3PAR Persistent Checksum, and measuring the storage system performance with both intentionally inserted bit errors and with unadulterated data, Demartek concluded. (See Figure 2.)

3PAR_Demartek_2.jpg

 Figure 2: Bandwidth of OLTP workload with Persistent Checksum enabled and disabled 

Demartek says in the report: “From this data, we conclude that for the workload evaluated, 3PAR Persistent Checksum has no net negative impact on either the amount of work done by the 3PAR StoreServ 8450 Storage array or the time in which it took to perform each I/O comprising that work.”

Demartek also measured response time by evaluating the time it took for the system to process the I/O requests.  Demartek concluded, “there was no statistically significant difference in average I/O latency between I/O running with Persistent Checksum or without.” (See Figure 3.)

3PAR_Demartek_3.jpg

 

 

 

 

 

 

Figure 3:  I/O latency with Persistent Checksum enabled and disabled

Why 3PAR Persistent Checksum is another feather in the 3PAR cap

When deploying all-flash systems, you are making an investment with the promise of better performance and response times, more data compacted in a smaller footprint and lower power consumption, to name a few. That being said, when it comes to data integrity, you should be asking yourself: Is my storage system adequately protecting my all-flash investment?

If it is that’s great, but it’s probably not the case, at least for midrange all-flash storage and some high-end storage whose T10Diff solutions often are restrictive and too dependent on the OS or application. T10Diff is a nice way to accomplish data integrity with flash and a FC fabric. But what is the cost premium and performance impact and flexibility of your T10Diff solution?  The Demartek evaluation of HPE 3PAR Persistent Checksum gave us insights into a standard HPE 3PAR StoreServ storage solution that is available across 3PAR midrange and high-end storage systems.  

Demartek demonstrated that Persistent Checksum effectively protects I/O throughout the SAN and that it virtually has no impact on storage bandwidth or I/O response times or array CPU utilization thanks to its ASIC offload. (See Figure 4.)

3PAR_Demartek_4.jpg

 

 

 

 

 

 

Figure 4

Finally, I want to leave you with a few interesting excerpts from the Demartek evaluation of 3PAR Persistent Checksum.

“Unlike some other storage vendors, HPE builds Persistent Checksum logic into its 3PAR StoreServ Storage array controllers, HPE 3PAR StoreServ 16Gb Fibre Channel Adapter as well as HPE StoreFabric Fibre Chanel Host Bus Adapters (HBAs) for media-to-host protection. Other vendors limit data integrity checking to high-end, Tier-1 hardware only while expecting operating system and application vendors to write complementary checksum verification code into their own products. By contrast, HPE brings the entire checksum process into the HPE 3PAR StoreServ Storage and HPE StoreFabric storage networking solution, thereby lowering risk and making this technology available to midrange storage customers as well.”

“How valuable is data, and what solution is best positioned to protect that data?” Mid-range, all-flash storage customers get an extra layer of protection by default with HPE 3PAR StoreServ Storage without sacrificing performance. And performance does drive purchasing decisions. We would consider this at least worthy of some consideration when evaluating upcoming storage acquisitions”

Read the complete report: Demartek Evaluation of HPE 3PAR Persistent Checksum

Check out the Demartek blog: Demartek Evaluation: HPE 3PAR Persistent Checksum

 

Ivan.jpg

Meet Around the Storage Block blogger Ivan Iannaccone, Director of Product Management - HPE 3PAR StoreServ Storage, HPE.  @IvanIannaccone

0 Kudos
About the Author

StorageExpert

Our team of Hewlett Packard Enterprise storage experts helps you to dive deep into relevant infrastructure topics.

Labels
Events
28-30 November
Madrid, Spain
Discover 2017 Madrid
Join us for Hewlett Packard Enterprise Discover 2017 Madrid, taking place 28-30 November at the Feria de Madrid Convention Center
Read more
HPE at Worldwide IT Conferences and Events -  2017
Learn about IT conferences and events  where Hewlett Packard Enterprise has a presence
Read more
View all