Migration of iSeries to Public Power Clouds using Commvault
Ivaylo Nikolov
in AS/400, Azure, Backup, Cloud Computing, IBM Cloud, iSeries, Migration, Skytap, Skytap on Azure
Intro
Over the years. we have been requested by customers and partners on many occasions what is the best approach to migrate iSeries workloads into the public Power-based clouds (mostly Skytap and IBM Cloud PowerVS). Obviously there are numerous approaches from an ongoing replication to physical transfer of data on media, which are not for here. However, one of the options that we have been intensively advising and implementing over the years is migration using Commvault backup software.
The option gets extremely approachable if we are also optimising the backup mechanisms of the customer and implementing a full-blown backup solution including compression, deduplication, integration with Object Storage solutions, secondary copies etc. Backup solutions are something we plan to cover in another article, this one will focus on migration approaches.
Commvault backup
We do presume that most of you are aware that Commvault is consistently a Leader in the Gartner Magic Quadrant for Enterprise Backup and Recovery Solutions, providing solutions across all platforms from endpoint devices to iSeries servers.
Due to its versatility, it is extremely useful when having different workloads, including in the Cloud where it integrates natively into the hypervisors. In the nutshell, we use Commvault as one of our prime tools for migrating data and tend to recommend it especially where we are asked about a backup solution including implementing resiliency for AirGap solutions.
iSeries networking
With iSeries itself, we are not going deep into its networking details, but we will speak a bit in general about how iSeries handles networking. Generally, iSeries is notoriously bad at handling anything that needs transfer of large amount of data over the network – especially when you have a tight migration window to meet. You can find a lot of articles suggesting that you mount some magic NFS share remotely and copy the data over it, or you use FTP to copy data over the network. From experience here, if you do not have days/weeks to migrate these environments, you better think of something else.
We often get involved in projects where teams have already struggled with this, so here are our suggestions based on experience (we fully respect that other specialist organisations like ourselves have developed other approaches) – this is just what has worked for us in complex scenarios where migration windows are tight.
Migrations to cloud
Now, usually when we talk to customers about their willingness to migrate to cloud, we do discuss the matter of the downtime windows. If you are working on very tight downtime schedule (I mean tight like minutes), forget about anything besides some kind of replication scenario like MIMIX, QuickEDD or equivalent (most such tools are owned by a single company), which are related to additional costs.
However, in most of the other cases, especially if you need a backup solution integrating into Azure Blob or IBM Cloud Object Storage, Commvault will do the job.
Why is it better to use Commvault than copy over the network? We will explain.
Commvault software is absolute monster in deduplicating and compressing the backup data. It compresses iSeries backup data so well that it basically tends to reduce the network traffic by up to 80%. In addition to that there will be deduplication advantages as well – not that big as the compression, but still very feasible. Additionally, Commvault works very well while copying data over WAN, which is essentially the case you are likely to have. Whether you are using Azure ExpressRoute, S2S VPN, or IBM Cloud Direct Link, Commvault’s WAN optimisation capabilities will help making copies stable and utilise the bandwidth at the best possible rate.
Of course, it sounds easier than it is as it needs configuration expertise to make it happen, but the end result is always excellent – migrating large amounts of data to meet tight windows.
Integration to Object Storage
The basic concept is that we use a single point of backup media in the Cloud, which is the Object storage itself. We are backing up source LPARs to the Blob/IBM COS and we are restoring from the same object storage to the cloud. This tends to really quicken the process and also these services are usually really suitable for uploading large amounts of data. Commvault is fully capable of deduplication on them and additionally it provides to you the possibility to integrate fully with immutable Blobs (AirGapped) and with geo-replication (easy second copy) as well.
Presence on-site
We always analyse and utilize the on-site environment. We set up on-site Commvault media agents close to the LPARs so we can backup quickly and we tend to optimise where we do compression and deduplication so we get the best possible speed from the backup. The media agents need to have enough resources to do multiple streams for multiple LPARs and utilise the available bandwidth to the cloud to the maximum. This needs to be monitored constantly, so for you to have expectations on the migration windows and on the utilisation of the resources.
Optimising network configurations
There are a number of settings that you need to do both on the iSeries LPARs and on the Commvault media agents, which tend to increase the backup speed sometimes 10 times even!!!
Be sure to take care of these – most of them are found in the public space as recommendations and can really do the difference. In a lot of cases, we have seen iSeries LPARs with ancient network configurations, so be aware and check this always. Things like TCP buffer sizes, LSO and MTU can bring a lot of improvement.
Optimising Commvault configurations
Last, but not least, you need to take care of Commvault configuration itself. Think about how you structure your largest libraries for backup, how you utilise data readers and application read size. Be careful with optimised scanning and while saving access paths. There are a number of things, which could save you hours of downtime if done right. So, if you need to do it alone, test and see, if not we can always help.
Restore tends to be the most important part for optimisation as the flow of restore is really important. If you mess up the order, you will need to do it again, unless you find out too late, which is even worse. Generally, Commvault has a default flow for migration of iSeries servers, but you will need to tweak it significantly to get the most of it.
At the end
Everything above might sound too complicated, but it is generally not, if you know what you are doing and have done it multiple times. So, if you cannot afford a migration partner like us (although we always try our best to fit shrinking budgets), make sure to test and see the results with different options that Commvault offers. Essentially that will help you optimise your backup and restore speed in future as well.
At the end, you should be able to migrate large chunks of data in limited time reliably with Commvault and public cloud Object Storage, whether to Skytap on Azure or IBM PowerVS.
Go ahead and contact us if you think we can do it together for you. We are aware of all the above tweaks…
