Flash Storage Is Table Stakes For Any IBM i Cloud
October 31, 2022 Andrew Johnson David Fahrenkrug
After a long wait, the IBM i market is finally entering the true cloud era, one where utility pricing, performance, flexible capacity, and managed services are all available for companies running the IBM i operating system and RPG, Java, PHP, and sometimes COBOL applications against the Db2 for i relational database.
But there are clouds, and then there are things that should not rightly be called a cloud.
And we think that if you are a modern cloud, then your primary storage has to be flash and that there is no place for a hard disk drive based on spinning rust anymore. But we also think you have to put your money where your mouth is and provide flash storage at the cost of disk storage, and in our case, moving to IBM’s FlashSystem FS7200 arrays from IBM Storwise V7000s with a mix of disk drives for capacity and flash drives for hot data is at least 50 percent more expensive to provide capacity – and we provide that flash capacity and all of the performance benefits at the same price.
The performance benefits of all-flash storage for IBM i customers alone are worth a premium, and we do not have any customers that are not seeing at least a 30 percent performance boost in moving to our cloud thanks in large part to our use of flash-only primary storage, and that is something that is built into our cloud architecture and that we do not charge a premium for. And we can do this in large part because we are buying FlashSystem FS7200s at volume, with nearly a petabyte in storage capacity and growing fast. (We have clustered storage from Infinidat that is used as primary storage for our X86 cloud.
When we moved from the V7000s to the FS7200s, we ate that cost and did not charge a premium for primary storage because we felt that it was our responsibility to keep up with the technology. It is our job to provide the best solution, and that is what we do.
But the move to all-flash on our cloud is not just about application performance. It is about being able to do all kinds of data management tasks at high speed as well.
For instance, we have created a set of custom scripting we call IntellaFlash so our customers can do no-downtime backups from their IBM i instances on our cloud. We find that customers use this IntellaFlash feature to create development and test environments that are refreshed from the production systems instantly. And for disaster recovery, using these no-downtime backups means we can create a backup and do a full system switch test of high availability clusters without actually taking the real production system offline.
The key to making these no-downtime backups usable in production environments is data compression, and for IBM i databases, customers typically see a 5:1 compression. If you have a lot of mixed data types in files on the Integrated File System, you might see the compression ratios drop to 3:1. And this is why this is important. As a matter of practice, when customers make no-downtime copies of data, we keep them out there on primary storage as long as they are being used and then we keep them around for 24 hours even after it is taken out of commission by customers just in case they need to revive them.
And to further secure the data in the flash storage, we use an open source virtual tape library product that runs on Linux from QuadStore Systems to do data de-duplication and push it out to our V7000 disk arrays, which we have kept around for this purpose. QuadStore supports IBM i and AIX backup as well as Linux and Windows Server environments, and also can emulate all kinds of tape drives and tape libraries from IBM, Hewlett Packard Enterprise, Quantum, ADIC, Overland Storage, StorageTek, and Spectra Logic.
Here is the last thing you need to consider when you choose a cloud for your IBM i workloads. We do not make use of any local storage on any of our Power Systems servers. Everything is running on the FlashSystem arrays. We have consolidated and secure storage, and maybe someday, we might have no local memory on our Power10 servers if IBM’s “memory inception” memory pooling works as advertised. We are getting our hands on a Power10 system now to test these ideas out, which are very interesting.
You deserve a cloud that stays on the cutting edge. Your business will depend upon it.
Andrew Johnson is chief technology officer and David Fahrenkrug is chief information officer at Focal Point Solutions Group.
This content was sponsored by Focal Point.