2021 saw the first full year of the new LTS/CD delivery mechanisms for software functionality updates within Spectrum Virtualize. With 8.4.0 initially being released this time last year, the LTS version is now classed as a recommended level and I discussed its contents, and some things to consider regarding pre-reqs, end of support etc in my previous post.
For 2021, the whole storage family got some new icons, Spectrum Virtualize is signified by the icon above.
1Q2021 – 184.108.40.206
We don’t normally add functional content to a PTF release, but with the release of the award winning, unique 1U, IBM NVMe FlashSystem 5200 in the first quarter, a lot of the groundwork for this had been already included in the 220.127.116.11 base level. Not many people pick up the .0 version and wanting to release the product on an LTS version, the 18.104.22.168 release in Feb included full support for the 5200 as well as the new GUI support for 3-site replication features (previously CLI only).
The 5200 was the first platform to gain advantages from the new inter-node messaging protocol. The old protocol was loosely based on a serial SCSI like system, with the new protocol being more parallel NVMe-like. This meant unlocking more potential when passing messages and data across the internal inter-canister PCIe link.
The 5200 also only supports our Distributed RAID (DRAID). Traditional RAID arrays are not supported on the 5200.
2Q2021 – 8.4.1
The first of three Continuous Delivery releases came in June with a round of up smaller “featurettes” that picked off some of the often requested feature enhancements.
Most notably the increase to the Volume Mirroring (VDM) sync rates. All Spectrum Virtualize based systems include a form of logical volume mirroring. Not to be confused with disaster recover, unless you are running an SVC stretched cluster (ESC) of course, VDM lets you have a single host volume that is made up of two volume copies. Those copies can be in the same or different storage pools, and in the SVC ESC case, in different sites or locations to provide High Availability.
The VDM function was originally introduced to all SVC to use “entry” level storage systems, with lower reliability rates and mirror the data between a pair of controllers, thus enhancing availability while keeping costs lower. These days its primarily used as a data migration method.
- Add a second copy to an existing volume (in a new pool, or controller)
- Let the existing data synchronise from the existing, to new copy
- Once both copies are in-sync, split off the original copy and delete it.
Using this method, we can non-disruptively migrate volumes from one storage pool/controller to another. The nice advantage this has over the pure “migrate” commands is that you can control the sync rate in terms of MB/s even stopping or pausing it if you need to. However the one major request here was to be able to set much higher sync rates. The feature was designed when the fastest storage we had was 15K RPM drives, so the 64MB/s limit was going to use about 2/3rd of the available drive performance. But with todays all flash solutions this simply needed increasing. With the introduction of 8.4.1 the sync rate can now be set to up to 2GB/s. Now with my old performance hat on – your mileage will vary – the sync rate is a “target” and so may not be achievable. In addition, ramp up the rate, as 2GB/s may kill your storage and cause unwanted performance issues to other volumes!
|Syncrate Value||Target Syncrate|
Additionally 8.4.1 completed the support for fully qualified domain names instead of just dotted decimal IP addresses for all features where you need to tell us about another machine, for example email servers, syslog servers etc.
Finally 8.4.1 also added support for Next Gen KeySecure and CipherTrust Manager for external encryption key management.
3Q2021 – 8.4.2
The second CD release for this year pulled in a very topical feature, IBM Safeguarded Copy. Unless you have been living under a rock in the last 2 years (and with COVID who would blame you) cyber crime is on the rise. Its almost weekly if not every other day that we hear about another large corporation being the target of a ransomware attack.
I will do some more posts about this in the new year, as the criminals are getting more astute, not only will the encrypt your data and probably your backups, but they will steal the data first and make money from it that way too!
Anyway, IBM Safeguarded Copy, and IBM CyberVault are a feature and a process by which you can easily recover should the worst happen. Recover is fine, but the average time to recover from a ransomware attack is 23 days… thats just to get the lights back on – with IBM Safeguarded Copy we can reduce that to hours, and with CyberVault that could be seconds or minutes.
Safeguarded Copy added in 8.4.2 allows the storage administrator to setup an immutable repository, on primary storage that allows the policy driven automation of snapshots (FlashCopy). You can decide the frequency and retention period and then the system automates the rest. Nobody to delete those copies, not even the storage administrator, and you cannot directly attach those copies to hosts, thus they are invisible to host based attacks, and protected against storage layer “bad actors”.
I highlighted “on primary storage” as this is key to being able to recover as quickly as possible. If you have your air-gap to tape or some other service, then its going to take considerable time to recover. With primary storage based snapshots, you can recover a copy instantly, validate it and start using it immediately!
CyberVault is around the processes you put in place to recover, making use of the Safeguarded copies, an isolated safe compute environment and the validation techniques you can use to ensure copies are not infected, or at least are usable pre encryption event. As I said, a whole topic in itself and I will cover more in the new year.
Some more features and featurettes added in 8.4.2 :
- Multiple IP addresses and VLANs per physical ethernet port
- Multiple IP partnerships – up to 4 – another long requested featurette
- Scalability improvements, HyperSwap (2K) volume count, Host Volumes (~16K), FlashCopy mappings (10K)
- Addressable capacity for HyperSwap and Replication increased to 2PB per IO group
- Provisioning Policies – define default volume creation attributes – useful for automation and RESTapi usage
- Volume Mobility – initial feature to allow migration of volumes in a non-disruptive manner between two disparate systems. This is another topic I will cover in detail in the new year.
- FlashCopy Volume expansion support
- The new Inter-canister NVMe-like protocol support on FlashSystem 7200 and 9200
4Q2021 – 8.4.3
The final CD release for the year is actually a cloud only release. It enhances the Spectrum Virtualize for Public Cloud (SV4PC) capabilities, bringing the software support up to include the features up to 8.4.2 but most notably now allows deployment on Microsoft Azure. You can deploy an environment much as with AWS using the Azure Marketplace and within minutes have a cloud based Spectrum Virtualize system ready to be logically configured with Microsoft block storage.
The same value propositions and use cases apply as before, reducing the cost of cloud block storage with thin provisioning and data reduction, deploying disaster recovery in the cloud, migrating workloads onto cloud, or migrating between cloud providers.
All in all, despite the global pandemic, its been a bumper year for Spectrum Virtualize features, picking off a few of the critical long requested features and at the same time, hats off to the development team who have mostly been remote working. Releasing new hardware platforms while most of the team works remotely is no small feat!
I hope to catch up with some of you face to face in 2022, and wish you all the best over the holiday season.