Long Distance Live Partition Mobility– A tale of collaboration

F5 Networks and IBM continue the long tradition of collaboration with the latest supported solution of Long Distance Live Partition Mobility. F5 worked closely with IBM to further prefect IBM Virtual I/O’s technology to better support Long Distance Mobility, proving again than when customers do business with F5 or IBM, they’re getting a wealth of value added benefits from the partnership.

WHAT IS IT?

Partition Mobility is IBM’s Power VM capability that allows for the transfer of active and inactive partitions from one V/IO server to another.  This solution has been offered since Power6 technology based systems, and in our testing, we were able to move a running machine between two datacenters separated by a simulated 1000 Kilometers.  The details of how we achieved this, a bit more on the solution is below, after the diagram.

 

SOLUTION OVERVIEW

The basics of the solution are that during Active Migrations, a running partition is moved from a primary LPAR (pictured above on the left) to a Failover LPAR (pictured above to the right).  Applications can continue to handle their normal workloads during this process.  The rest of the picture is comprised of these pieces: The IBM Hardware Managed Console (HMC - pictured above to the top left), the IBM Integration Virtualization Manager (IVM – not pictured), the shared storage (pictured in both data centers) and the F5 BIG-IP technology that enables this, specifically, F5 BIG-IP WAN Optimization Module to enable, secure and accelerate the data transfers and F5-BIP EtherIP, to keep active client sessions connected during transfers and finally, F5 Global Traffic Manager (GTM – not pictured and optional) to direct incoming traffic intelligently during failover events.

All of the details of about IBMs Power Mobility feature can be found here: http://www.redbooks.ibm.com/redbooks/pdfs/sg247460.pdf

Recommended reading from F5 about setting up these environments can be found here: http://www.f5.com/pdf/deployment-guides/f5-vmotion-flexcache-dg.pdf and http://www.f5.com/pdf/white-papers/cloud-vmotion-f5-wp.pdf

We’ve built this solution in lab but a deployment guide is still pending, so in the meantime, I hope that my VMotion deployment guide and the white paper on cloud migration will fill in any questions you have about the nuts and bolts of the deployment.  The solutions are very similar while, of course, the under lying technologies are unique to each partner. You can of course email me with any questions you have as well, at n dot moshiri at f5.com

WHAT IS REQUIRED?

The basics are as follows:

• AIX 7.1 is recommended,
• VIOS Version 2.2.2.0 (release August 2012) is recommended,
• Storage with connectivity to both data centers (or tightly couple replication),
• A latency (and distance) between the two data centers that can support the nature of the application running in the infrastructure
• And of course, network connectivity between the two data centers.

I will throw in a quick word about VIOS release 2.2.2.0.  During initial testing we discovered that IBM’s mobility manager picked an arbitrary TCP port during migration events.  For internal migrations this would pose little problem, however, in order to secure, optimize and allow transmission over firewalls, in the long distance scenario, arbitrary ports would simply not do.  IBM stepped up and delivered.  With version 2.2.2.0, a user selectable port range allows for migration events to happen in a much more controlled manner on the network.

Bottom line, migration traffic between the two data centers need to be secured, accelerated and client traffic needs to stay up and know where to go. BIG-IP provides all of this functionality through Local Traffic Manager (LTM), WAN Optimization (WOM) and Global Traffic Manager (GTM).

This can be an ideal solution for certain use cases.  When examining these architectures analyze:

• Network connectivity between the data centers,
• Shared Storage between the data centers,
• What are the workloads on the partitions, what are the memory footprints.

Every Partition Mobility architecture will be different.  Reach out to me or your F5 FSE and definitely plan on several rounds of architectural review.  F5 and IBM will be there to back you up.