Hello again my virtual friends.
Now when EMC world is behind I found the time for some tech updates. This time, is about another solution we have tested leveraging EMC’s VPLEX Geo.
VPLEX is a solution for federating EMC/Others storage. it sits between the servers and your storage adding a virtualization layer. but it can do much more than that as it presents a sophisticated SDRAM cache which can be distributed to a remote site while maintaining coherence. EMC VPLEX family consists of three viable configurations/offerings:
- VPLEX Local – For managing data mobility and access within the walls of your data center using a single VPLEX cluster
- VPLEX Metro – For mobility across two sites separated by an inter-site latency of up to 5 ms (roundtrip). We have tested that solution last year; vMotion over distance for Microsoft, Oracle and SAP.
- VPLEX Geo – For access between two sites over extended asynchronous distances with up to 50 ms latency (RTT).
Couple of months ago the virtualization team in Hopkinton worked on testing application mobility on VPLEX Geo this time with Microsoft Hyper-V clustering (sounds strange?! yes, we work with both VMware and Hyper-V in our labs but don’t expect us to add a NetApp FAS or anything crazy like that 🙂
Its an interesting whitepaper that covers SAP, Oracle and SharePoint mobility with heterogeneous VMAX and VNX configuration. the wan link was optimized using EMC’s Select parnter Silver Peak but let me highlight the SharePoint part in that solution. for further reading please download Long distance application mobility – Enabled by VPLEX Geo.
The SharePoint farm had few site collecions and a total of 400GB of user content. a total of seven VMs constituted the server farm – 3 WFE, 2 Index/Crawl, 1 App and 1 SQL. the configuration supported more than 12,000 users with 10% concurrency with a sub-3sec user response time for all operations (browse, search, modify). Hyper-V clustering was configured using CSVs (cluster shared volumes).
The highlight of the test was migrating the ENTIRE farm from site A to site B with simulated distance of 2,000 km (~1200 miles) WITHOUT a disruption of service using Live migration. while migration took place, the farm’s user load capacity somewhat degraded but still was able sustain more than 9,000 users (-23% load).
Using Silver Peak WAN optimization in that solution resulted in almost 70% reduction of data transferred between the two sites. impressive!
To summarize, you have a virtual infrastructure stretched across 2 sites with ZERO downtime for migration operation and very low downtime for outage/disaster scenarios while the server/hypervisor infrastructure addresses the storage element as a SINGLE entity.
Stay tuned… best practices for SharePoint is coming in my next post.
E!
Leave a comment