One of the fastest growing companies in Texas, Drilling Info, Inc. provides drilling and other data along with analytical tools and applications to over 3000 companies that together account for over 90% of U.S. oil and gas produced. After evaluating several options from traditional server manufacturers for hosting his VMware based virtual infrastructure, Drillinginfo IT architect, Mark Domel, chose Cisco UCS.
"In a typical blade solution," said Mark, "we would have needed to add all these points of management each time we added a blade chassis. We would have had to manage each switch, each chassis and also the SAN fabric by extending it onto yet another fibre channel switch. UCS gives us the flexibility to deploy additional chasses as needed without having to add additional points of management and without needing to configure network and storage switches."
Unlike traditional server products, Cisco UCS was designed from the ground up as a new type of optimized hosting platform for virtual infrastructure. Mark is confident in Cisco's abilities, "Cisco has been engineering and deploying modular and blade solutions for nearly 20 years. They're not behind the curve; they're way ahead of it. Factor in Cisco's history of pioneering technology and you've got a solid solution that's easy for IT architects to believe in. The way I see it is that UCS is about building a total solution, not just a response to a specific datacenter hardware need. Pair up matched virtualization and storage technology and you've got everything you need in a datacenter."
Drillinginfo Virtual Infrastructure Schematic
Mark and the Drillinginfo IT team compared various VMware vSphere host options including Dell Blades, Dell Rackmounts, HP Blades, HP Rackmounts, IBM Blades, IBM Rackmounts and Cisco UCS Blades. He summarized, "As the apples-to-apples quotes starting coming in, we noticed that the UCS solution was significantly less expensive than the alternatives. Not only was the pricing more aggressive, but it required far less of an investment in the networking components as they were already included in the design. Also, UCS utilizes a 10Gb network as its base architecture. This gave it a significant advantage from a future proofing and simplicity of design standpoint."
The Dell Rackmount solution, while considerably more expensive than the Cisco UCS, was the most competitive of the non-UCS quotes in both price and rack space consumed. It allowed Drillinginfo to use the four onboard NICs with ISCSI TOE for the SAN and an add-in quad-port NIC for the LAN, although Mark said, "the cabling would still be a steep administrative burden."
Other Key Factors in the Decision Process
The competitive up-front cost of the UCS was just one of the many variables weighing in Mark's decision process. Other key points included cabling, implementation ease, growth, break-fix, administration, compatibility, supportability and size.
Cabling: "There's this great image floating around the Internet of a comparison between standard servers, a blade center, and UCS. It's really very telling. Having worked in a very diverse and segregated network environment and deployed both high density traditional servers as well as blade solutions I can tell you that neither are fun from a datacenter administrator's point of view."
Implementation: "In our case the time it took to un-box the solution, rack, cable, bring online, and install VMware was about six hours with two guys. We were building VMs and remarking at how fast it all went. Planning and managing cabling from a traditional blade solution to the storage and Ethernet networks is a usually a major task. However, with UCS it was all just as simple as choosing how much redundancy and performance we wanted between the 6120s and the chassis and then connecting the uplinks to the network and storage."
Growth: "It was very important to me that we have a solution that's simple to grow. UCS embodies that in a way none of the other solutions could. Adding another chassis is as simple as racking it and plugging in the interconnect and power cables. That's it! You don't have to configure the LAN, SAN, or individual chassis modules. Just plug it into the UCS fabric and you're off to the races, amazing. Additionally, I have pools configured such that if a blade server of a certain hardware specification is inserted UCS will automatically configure it to be a VMware host with WWNN, WWPN, MAC, and naming from pools. Being able to pre-provision storage and network security devices based on hardware id pools is just awesome."
Break-fix: "Replacement of a blade no longer means getting involved with anything other than simply replacing the gear. We can have the replacement blade shipped to our collocation provider, we put the blade in maintenance mode, they remove the defective, insert the replacement, and we assign that replacement blade to the defective's profile. UCS will automatically give that replacement blade the hardware identifiers associated with that profile. Simply put, UCS abstracts the hardware from the server. This requires boot from SAN, which is incredibly convenient."
Administration: "The UCS Manager interface is far and away the best of breed interface I've used. In comparing it to the Dell, HP, and IBM systems I've managed in the past, I certainly have developed a love for how logically it's laid out. A lot of what you'll pay extra for to buy something like IBM's Open Fabric Manager is included in the UCS solution – multi-chassis blade redundancy for example. It doesn't require a separate server and is already redundant as the solution is designed that way. UCS feels to me like it's more IT administrator centric than the counterparts that I've used."
Compatibility: "Microsoft guest operating systems on VMware virtualization on Cisco UCS blade servers with EMC storage generated the perfect union of compatibility for us. Every vendor involved is aware of our solution and accepts that it all works perfectly together."
Supportability: "Both EMC and VMware have support facilities with which most IT personnel are familiar. Cisco's TAC is pretty unique to network and security administrators. Well sys admins rejoice, you don't know how good the network guys have had it. TAC is a true gem in the support realm. Those guys are all-stars technically and Cisco's support model is the best I've encountered. Throughout my time as a network engineer I've always had comfort in knowing that TAC had my back if something hit the fan. I'm now so glad to see that as an architect I can put a solution in place that gives our team a strong safety net. When we call TAC about UCS, we do so knowing that that we'll be well taken care of despite the many technical disciplines that might be required to troubleshoot and resolve the issue. The typical TAC guy is going to be far more versed in Ethernet networking, storage networking and virtualization than when calling a traditional server manufacturer. I certainly don't trust any other server vendors to jump into the CLI of my production SAN switches."
Size: "Size is everything. Relatively speaking, we're a small shop. I couldn't commit to a solution designed to start at massive scale. We needed something that could start off just providing basic redundancy and scale to meet our needs over the next five years of growth. UCS fits the bill. There are tons of presentations showing how big it gets but I think Cisco sometimes forgets to show how small it works too."
Savings so Far
Drillinginfo was able to reduce its footprint from five racks down to two, saving nearly three-thousand dollars a month in collocation costs. Other cost savings result from decreased Microsoft Windows Server and SQL Server licensing required under virtualization. About 70 VMs currently run on the three vSphere UCS blades, and according to Mark, "they're dramatically outperforming their physical counterparts. The savings in power consumption alone is enough to pay for those hosts and licensing within a year. We chose to install the SQL servers directly on the blades to take advantage of the full memory, CPU, and IO capabilities of the systems. We've reduced our licensing costs there dramatically by cutting the number of sockets by 75% and increased performance by 400%."
"So far we've seen an average of about four times performance increase by using virtualization on the new Intel Nehalem-class processors when compared to the two and three year old physical systems. This has reduced the quantity of our operating systems and applications by virtue of the new VM systems outperforming the old model, saving licensing dollars and IT administrative costs.
"Although I'm not aware of the numbers, I know we've saved monies in our travel budget by not having to reimburse admins for trips to the collocation facility. Between UCS Manager and VMware vCenter we're able to do just about everything remotely."
Future Drillinginfo Virtualization Initiatives
In addition to evaluating the potential virtualization of its desktop environment, Drillinginfo is also increasingly enabling cloud access to its customers in order to access data in a raw format. The company is considering eventually making VMware Lab Manager virtual machines available to customers to help them build their own environments. The UCS plays an important part in enabling its cloud activities by providing great flexibility for multi-tenancy environments. According to Mark, "…it lets the IT staff pre-provision a UCS slot to match a pool which is then automatically configured for the ESX host to provision storage, etc."
Mark believes that the success of the UCS will spur other server manufactures to step up their games. From this perspective, he compares the UCS with Apple's iPhone, "The iPhone package Apple put forward was so innovative, well integrated and high quality that everyone else had to take it up a notch. The same thing will happen with the virtualization market and everyone will benefit from Cisco's hard work."
Drillinginfo's virtual infrastructure environment includes:
2x Cisco UCS 6120XP 20-port Fabric Interconnects
2x Cisco UCS 5108 Chassis
2x Cisco UCS B250 M1 blades w/ 2x Intel x5550 CPU, 48x 4GB PC3-10600 RAM, 1x Emulex M71KR-E 10Gb CNA
1x Cisco UCS B250 M2 blade w/ 2x Intel x5650 CPU, 48x 4GB PC3-10600 RAM, 1x Emulex M71KR-E 10Gb CNA
3x Cisco UCS B200 M1 blades w/ 2x Intel x5550 CPU, 12x 4GB PC3-10600 RAM, 1x Emulex M71KR-E 10Gb CNA
2 x Cisco UCS B200 M1 blades w/ 2x Intel e5520 CPU, 12x 4GB PC3-10600 RAM, 1x Emulex M71KR-E 10Gb CNA
2x Cisco Catalyst 3560E-48 Ethernet switches
2x Cisco MDS 9124 Fibre-channel switches
1x EMC Celerra NS120
- 2x Celerra NS40 Data Movers
- 1x CX4-120 w/ 2x 300GB FC DAE, 2x 1TB SATA DAE, 1x 67GB SSD DAE
VMware Enterprise Plus
Cisco Nexus 1000v
Microsoft Server 2008 R2 Datacenter
Microsoft SQL Server 2008 Enterprise
Ubuntu Server 9.10 x64
Thanks to Brent Blaha and the INX Austin team for introducing me to Mark.