Previously, the physical trappings of Cisco UCS, Intel chip-sets and UCS Manager were the most useful part of this server system. As we embrace new Intel CPU's, Chip-sets and memory, we are gaining added value from the original UCS design - which was a software construct based on XML API's and a suite of code that is really starting to blossom as a central automation vehicle, that scales to deliver new features and extended integration with a suite of security, management and performance offerings Cisco has added to its portfolio.
While UCS hardware leveraged standard x86 designs, the use of a single pane of glass to monitor, deploy and provision servers was a huge timesaver. Since its UCS release in 2009, Cisco has extended the core functionality with Central, a tool for managing multiple domains, Director - an automation tool and Performance Manager. In the past few years CIsco has been on a buying binge for the Data Center, snapping up Cliqr, Lancope, AppDynamics, ContainrX, and several others that are being integrated with in-house analytics tools like Tetration and external tools like Turbonomic to provide an incredibly powerful, secure automation platform that will be the foundation of a future autonomic server environment with adaptive security and dynamic self diagnosis.
Cisco UCS Manager is embedded in the cost of the fabric controllers and is used to manage the servers, chassis and fabric. It also serves as a link point for integrating tools like Director, Performance Manager and Central. Future additions to the UCS tool set are extensions that Cisco is feeling out how best to offer to customers - for straight purchase - or via subscription.
I encourage UCS users and those considering UCS adoption to dig into the subscription offerings and get some clarity on how they grow over time. For example, as powerful security tools like Stealthwatch (Lancope) are added, what other systems are required and how are those subscriptions managed. When Analytics are required - do you need a Gigabuck Third Party offering or are you going to jump on Cisco's Tetration bandwagon and roll your own? I push for simplicity with Cisco. However, you need good data for that conversation. Talk to the apps, dev and ops teams as to what is needed today, where you are going and what future needs will become vs what might be nice to have. Once you understand where you are going, you are in a much better position to negotiate with a relative newbie like Cisco on how best to get there.
Things will only get better going forward. UCS Manager is an XML construct. Everything is in software and can scale and expand with increased hardware capability, while other architectures require extensive effort on each end to develop hardware, then update and test a new rev of software for reliability and consistency.
The big challenge for Cisco today is they built UCS manager for Cisco CCIE's anxious and able to have every knob and dial available to tweak. As a result, UCS manager is overly complex relative to functions and features and a lot of effort can go into streamlining and simplifying the User Experience. However, after 8 years in the market and huge acceptance of its increased ROI over competitive offerings and an appreciation for what UCS provides in OPex reduction, you can buy experienced UCS engineers vs having to develop and train them, only to see them purchased by a competitor.
Improvements to My Organization
I have a client who is currently managing 1500 servers with two people for a mission-critical retail operation. Previous operations teams using HP and IBM servers required 4x more people to manage the same number of servers.
Room for Improvement
This product comes from Cisco, who is fourth in the worldwide supply chain. That means options take a bit longer to get to their platform, as they insist on doing their own quality validations. Right now, the market is rapidly transitioning to solid-state media and the Cisco options tend to be less varied and more expensive than a broader slate of products from HP, Dell or IBM.
Cisco UCS offers a scalable platform with tremendous OpEx advantages. However, Cisco does not have the storage play that Dell (With Cisco Partner EMC in its fold) and HP have. With their long position in the market place on the PC supply chain side, both Dell and HP source and deliver high volume, low cost, advanced enterprise solutions from previous consumer focused suppliers like Samsung and Toshiba. Example’s like Sandisk’s 3.8TB SSD used in EMC VxRail products and newly announced Samsung 15TB and 6.4TB 1M IOPs SSD come to mind. While Cisco still carries the earlier versions of similar technology from FusionIO, the next gen lower cost options from Samsung will take a while to be approved and provided by Cisco.
Cisco’s internal testing and validation processes – to assure UCS Manager compatibility - mean they lag both HP and Dell in delivery on the newest storage paradigms – specifically the breadth of the SSD and NVRam offerings. Both these trends (High performance, High capacity SSD, and NVRam) offer major changes in architectural models. For organizations that seels to push the bleeding edge in testing and development, UCS will lag in delivery by a quarter or two. This has little impact on mainstream enterprises who will not adopt before a technology is thoroughly vetted by industry “Pioneers” – usually mid-sized shops that “took a chance” on introducing a new platform into their relatively modest environment.
Use of Solution
I have used UCS since 2008, when the product was first released.
No issues with stability that we have not seen across other systems. In particular, due to Cisco networking dominance, the focus is on drivers that work with their products for all the competitors as well. Networking is typically the server area with the most work to be done – but this is the strength of Cisco.
UCS originally promised to support 40 chassis per fabric – that has now been scaled down to 20 – which limits users to domains of “just 160” physical server blades. This has not proven to be an issue or obstacle. The release of UCS Central provides software to manage an array of fabrics so we can scale to thousands of physical servers.
Customer Service and Technical Support
This is a foundational core of the Cisco Data Center automation experience and is a far more robust platform than currently provided by competitors. Customer service from Cisco and its partner community is thus on par with the same exemplary service provided by its TAC teams for business critical network deployments. Technical Support
Leveraging Cisco Network Technical Monitoring – the ability to call for a case and get resolution - is a process we are well aware of and very comfortable with.
HP was the incumbent, displaced by UCS, which has proven easier to manage scale and use. The HP system just had too many pieces and the iLO lockin was a major cost that the UCS architecture leapfrogged.
Initial setup requires some training due to its scale. It’s like riding a car vs driving a truck. You use the auto driving skills when you drive a truck – but there are a few things to be aware of. One of the nuances with UCS is that it is a fully abstracted, scalable environment. So you can set up your domain to accommodate a single server or 160 servers. This requires adopting a standard naming convention, IP addressing, etc. Once those are established, like a truck vs a car – you can haul a lot more freight with UCS.
Pricing, Setup Cost and Licensing
Obviously, the worst-kept secret with all vendors is to negotiate as close to fiscal year-end as possible. For Cisco, the year-end is July 31st, so they are well positioned for organizations deploying summer projects. The other issue is the move to bundle licenses. That is great for highly dense environments like a data center, but it makes much more sense for individual licenses for distributed environments like hundreds of storefronts or clinics distributed across a wide geographic area.
Other Solutions Considered
As stated earlier, we had HP. As a marquis client they fought hard on equipment price to maintain their position. However, the decision was based on OpEx, which greatly favored UCS. Once we had a few systems in place and people trained up on their use, it was not long before HP was displaced. Because both the IBM and Dell management architectures were quite similar, we looked and got a few quotes, but did not see anything to justify further evaluation resources.
The biggest issue is automation. How to move the mundane tasks from people to machines; alert filters to improve management productivity and reduce overhead. Cisco is deploying a suite of products (Central, Director, Performance Manager, etc.), as are IBM, HP and even Dell. However, UCS manger provides such a robust base that the ability to scale and realize benefits is greater.
At the end of the day, the UCS product requires planning before just jumping in, due to its ability to scale. As a user, you need to evaluate naming conventions, IP addressing models and so forth – think about the entire enterprise as opposed to a single server or rack of servers.
Use very good hardware and innovative network elements, such as the VIC 10Gb cards that allow for traffic sequestration and partitioning across multiple virtual channels in a single link and of course UCS Manager. I actually have the patent on similar IP when we started blade server systems with an acquisition by Intel. The direct spin-off was the IBM Blade Center, but due to the IBM investment in Tivoli, they never used our central management system. Cisco took a network- vs compute-centric perspective as they embarked on their server designs and, with a clean sheet of paper, evolved a centralized manager for deployment and systems management that enables huge scales in management productivity.
Disclosure: I am a real user, and this review is based on my own experience and opinions.
Jun 23 2017