Avatar
Evolving Cisco UCS for Big Data and the Internet of Everything
Image credit: imediaconnection.com

In an earlier blog, I discussed the incredible success behind the Cisco Unified Computing System (UCS) in Darwinian evolutionary terms.  Since I wrote that blog three months ago, we’ve continued to grow rapidly and have strengthened our position as the #2 blade server player worldwide from 19.3% to 21.5% revenue share (per IDC 2CQ2013 server market tracker, September 2013), with 33.9% revenue share in the US.

Prediction – The #1 spot is well within our reach sooner than you may think.

As we grow our installed base with roughly 1,000 new customers every month, our conversations about the future of UCS have taken an interesting turn. Until now in what I’ll call the “UCS 1.0” phase, Cisco focused on virtualization and private cloud as the dominant use cases that were top-of-mind for industry CIOs and we struck a resonant chord based on our growth – just look at our numbers.

We were market makers with expanded-memory 2-socket Intel EX blades (remember the B230’s?), which were gobbled up like candy into large-scale VDI deployments much to the surprise of the industry. We also jump-started a very attractive RISC-to-x86 migration practice, including Cisco IT’s own production environment: a 40TB mission-critical database that ran on HP Superdomes – a “circle of life” moment for me since Superdome was my program from 1999-2003.

We’ll continue leading in customer value for our original design centers, but we are now focusing on market expansion with what we call “UCS 2.0”, expanding into data-intensive, mission critical, analytics and service provider cloud environments with an increased level of R&D funding and strong corporate support from our top executives.

Prediction – You’ll see us more focused on architectural solutions for key industry vertical markets with tuned solution environments that leverage Cisco’s wide portfolio and that of our partners.

One such act of support is the announcement today of our intent to acquire WHIPTAIL, a leading solid-state systems company that boasts the highest scalability in performance and capacity of any scale-out flash vendor on the market today.  WHIPTAIL systems span from single-node entry products to 30-node behemoths that drive almost 400TB’s of flash, 40GB/sec of bandwidth and 4 million random R/W IOPS – for starters.

Prediction: Cisco will unseat Infiniband with low-latency Ethernet fabrics. Check out our USNIC technology for starters…

In our customer interactions it became very clear they view application acceleration using persistent solid-state memory as a use case that belongs in the server tier, not the storage tier.

In an application-centric world, we started thinking not about server vs. storage infrastructure, but how applications viewed data – hot “important right now” data, warm “may be of interest data” and cold “let’s keep it around for background mining or compliance” data.

We arrived at the conclusion that UCS needed to be best-in-class at accelerating hot data layers. Hot data is closest to applications and therefore has high affinity for the server tier.  Hence WHIPTAIL.

Assertion: Flash is a “boundary technology” that can be viewed as part of the memory or storage hierarchy. With respect to storage it’s faster and more expensive per GB. With respect to DRAM memory it’s slower but cheaper per GB. It therefore allows cost/performance arbitrage for applications by applying an accelerated persistent data model that can save on DRAM and de-complicate underlying permanent backing stores.

WHIPTAIL is a great fit with the fabric computing UCS architecture and also complementary to our C-Series rack mount servers and our SingleConnect capability in our UCS Manager that allows mixed-density blade/rack deployments to be managed from a common pane of glass.

Our intent is to fully integrate UCS computing and WHIPTAIL solid-state technologies over a Nexus fabric to create scalable persistent memory systems. That’s our vision.

Why? Because customers will be able to do things they could not before. Such as loading vast amounts of data in seconds and minutes, not hours or days – or – shrinking their performance footprint to a rack vs. 30 racks – or – accelerating Hadoop on all solid-state infrastructure – or – extending in-memory analytics to a scale previously not thought possible. That’s why.

As converged infrastructure advances as an ensemble computing architecture, boundary technologies like solid-state memory can be viewed as part of the memory or the storage hierarchy. Cisco’s point of view is to make it part of the memory hierarchy in the compute tier. That allows customers the best of both worlds – performance acceleration for applications while retaining their investment in permanent backing stores and simplifying their overall data center total cost of ownership (TCO).

To close on a Darwinian note, if UCS existed in the Cretaceous Period it would have been a Velociraptor  (meaning ‘swift seizer’)– sleek, fast and ferocious – eating everything in its path.  Velociraptors are believed to have hunted in packs, which is great considering the strong partner ecosystem that Cisco and UCS have built with industry leaders like EMC, NetApp and VCE as shining examples. We are committed to maintaining and expanding our hunting pack – more on that later!

if Cisco UCS existed in the Cretaceous Period it would have been a Velociraptor
if Cisco UCS existed in the Cretaceous Period it would have been a Velociraptor. Image credit: dark.pozadia.org


Authors

Paul Perez

Senior Vice President, General Manager

Chief Technology Officer, Data Center Group