The Little Mainframe That Could
RFG Perspective: The just-launched IBM Corp. zEnterprise BC12 servers are very competitive mainframes that should be attractive to organizations with revenues in excess of, or expanding to, $100 million. The entry level mainframes that replace last generation's z114 series can consolidate up to 40 virtual servers per core or up to 520 in a single footprint for as low as $1.00 per day per virtual server. RFG projects that the zBC12 ecosystem could be up to 50 percent less expensive than comparable all-x86 distributed environments. IT executives running Java or Linux applications or eager to eliminate duplicative shared-nothing databases should evaluate the zBC12 ecosystem to see if the platform can best meet business and technology requirements.
Contrary to public opinion (and that of competitive hardware vendors) the mainframe is not dead, nor is it dying. In the last 12 months the zEnterprise mainframe servers have extended growth performance for the tenth straight year, according to IBM. The latest MIPS (millions of instructions per second) installed base jumped 23 percent year-over-year and revenues jumped 10 percent. There have been 210 new accounts since the zEnterprise launch as well as 195 zBX units shipped. More than 25 percent of all MIPS are IFLs, specialty engines that run Linux only, and three-fourths of the top 100 zEnterprise customers have IFLs installed. The ISV base continues to grow with more than 7,400 applications available and more than 1,000 schools in 67 countries participate in the IBM Academic Initiative for System z. This is not a dying platform but one gaining ground in an overall stagnant server market. The new zBC12 will enable the mainframe platform to grow further and expand into lower-end markets.
zBC12 Basics
The zBC12 is faster than the z114, using a 4.2GHz 64-bit processor and has twice the maximum memory of the z114 at 498 GB. The zBC12 can be leased starting at $1,965 a month, depending upon the enterprise's credit worthiness, or it can be purchased starting at $75,000. RFG has done multiple TCO studies on the zEnterprise Enterprise Class server ecosystems and estimates the zBC12 ecosystem could be 50 percent less expensive than x86 distributive environments having the equivalent computing power.
On the analytics side, the zBC12 offers the IBM DB2 Analytics Accelerator that IBM says offers significantly faster performance for workloads such as Cognos and SPSS analytics. The zBC12 also attaches to Netezza and PureData for Analytics appliances for integrated, real-time operational analytics.
Cloud, Linux and Other Plays
On the cloud front, IBM is a key contributor to OpenStack, an open and scalable operating system for private and public clouds. OpenStack was initially developed by RackSpace Holdings and currently has a community of more than 190 companies supporting it including Dell Inc., Hewlett-Packard Co. (HP), IBM, and Red Hat Inc. IBM has also added its z/VM Hypervisor and z/VM Operating System APIs for use with OpenStack. By using this framework, public cloud service providers and organizations building out their own private clouds can benefit from zEnterprise advantages such as availability, reliability, scalability, security and costs.
As stated above, Linux now accounts for more than 25 percent of all System z workloads, which can run on zEnterprise systems with IFLs or on a Linux-only system. The standalone Enterprise Linux Server (ELS) uses the z/VM virtualization hypervisor and has available more than 3,000 tested Linux applications. IBM provides a number of specially-priced zEnterprise Solution Editions, including the Cloud-Ready for Linux on System z, which turns the mainframe into an Infrastructure-as-a-Service (IaaS) platform. Additionally, the zBC12 comes with EAL5+ security, which satisfies the high levels of protection on a commercial server.
The zBC12 is an ideal candidate for mid-market companies to act as the primary data server platform. RFG believes organizations will save up to 50 percent of their IT ecosystem costs if the mainframe handles all the data serving, since it provides a shared-everything data storage environment. Distributed computing platforms are designed for shared-nothing data storage, which means duplicate databases must be created for each application running in parallel. Thus, if there are a dozen applications using the customer database, then there are 12 copies of the customer file in use simultaneously. These must be kept in sync as best as possible. The costs for all the additional storage and administration can make the distributed solution more costly than the zBC12 for companies with revenues in excess of $100 million. IT executives can architect the systems as ELS only or with a mainframe central processor, IFLs and zBX for Microsoft Corp. Windows applications, depending on the configuration needs.
Summary
The mainframe myths have misled business and IT executives into believing mainframes are expensive and outdated, and led to higher data center costs and sub-optimization for mid-market and larger companies. With the new zEnterprise BC12 IBM has an effective server platform that can counter the myths and provide IT executives with a solution that will help companies contain costs, become more competitive, and assist with a transformation to a consumption-based usage model.
RFG POV: Each server platform is architected to execute certain types of application workloads well. The BC12 is an excellent server solution for applications requiring high availability, reliability, resiliency, scalability, and security. The mainframe handles mixed workloads well, is best of breed at data serving, and can excel in cross-platform management and performance using its IFLs and zBX processors. IT executives should consider the BC12 when evaluating platform choices for analytics, data serving, packaged enterprise applications such as CRM and ERP systems, and Web serving environments.
Disruptive Changes
Lead Analyst: Cal Braunstein
Amazon Inc. and Microsoft Corp. lowered their pricing for certain cloud offerings in attempts to maintain leadership and/or preserve customers. Similarly, Hewlett-Packard Co. (HP) launched its next-generation Moonshot hyperscale servers. Meanwhile, IDG Connect, the demand generation division of International Data Group (IDG), released its survey findings that show there may be a skills shortage when it comes to the soft skills required when communicating beyond the IT walls.
Focal Points:
- Earlier this month Amazon price reduced the prices it charged for its Windows on-demand servers by up to 26 percent. This brought its pricings within pennies of Microsoft's Windows Azure cloud fees. The price reductions apply across Amazon's standard (m1), second-generation standard (m3), high-memory (m2), and high-CPU (c1) instance families. CEO Jeff Bezos stated in the Amazon annual report the strategy of cutting prices before the company needs to, and developing technologies before there is a financially motivating factor, is what protects the company from unexpected markets shifts. Microsoft has responded by aggressively cutting its prices by 21 to 33 percent for hosting and processing customer online data. In order for customers to qualify for the cuts they must make monthly commitments to Azure for either six or 12 months. Microsoft also is making its Windows Azure Virtual Network technology (codenamed "Brooklyn") generally available effective April 16. Windows Azure Virtual Network is designed to allow companies to extend their networks by enabling secure site-to-site VPN connectivity between the enterprise and the Windows Azure cloud.
- HP launched its initial Moonshot servers, which use Intel Corp. Atom low-cost, low-energy microprocessors, This next-generation of servers is the first wave of hyperscale software defined server computing models to be offered by HP. These particular servers are designed to be used in dedicated hosting and Web front end environments. The company stated that two more "leaps" will be out this year that will be targeted to handle other specific workloads. HP claims its architecture can scale 10:1 over existing offerings while providing eight times the efficiency. The Moonshot 1500 uses Intel Atom S1200 microprocessors, utilizes a 4.3U (7.5 inch tall) chassis that hosts 45 "Gemini" server cartridges, and up to 1800 quad-core servers will fit into a 42U rack. Other x86 chips from Advanced Micro Devices Inc. (AMD), plus ARM processors from Calxeda Inc., Texas Instruments Inc., and Applied Micro Circuits Corp. (AMCC) are also expected to be available in the "Gemini" cartridge form factor. The first Moonshot servers support Linux, but are compatible with Windows, VMware and traditional enterprise applications. Pricing starts at $61,875 for the enclosure, 45 HP ProLiant Moonshot servers and an integrated switch, according to HP officials. (For more on this topic see this week's Research Note "HP's Moonshot – the Launch.")
- According to a new study by IDG Connect, 83 percent of European respondents believe there is no IT skills shortage while 93 percent of U.S. respondents definitely feel there is a gap between the technical skills IT staff possess and the skills needed by the respondents' companies. IDG attributes this glaring differentiation to what are loosely defined as "hard" (true technical skills and competencies) and "soft" (business, behavioral, communications, and interpersonal) skills. The European respondents focused on hard skills while their American counterparts were more concerned about the soft skills, which will become more prevalent within IT as it goes through a transformation to support the next-generation data center environments and greater integration with the business. As IT becomes more integrated with the business and operational skill requirements shift, IDG concludes "companies can only be as good as the individuals that work within them. People … are capable of creative leaps of thinking and greatness that surpass all machines. This means that any discussion on IT skills, and any decision on the qualities required for future progression are fundamental to innovation. This is especially true in IT, where the role of the CIO is rapidly expanding within the enterprise and the department as a whole is becoming increasingly important to the entire business. It seems IT is forever teetering on the brink of bigger and better things - and it is up to the people within it to maximize this potential."
RFG POV: IT always exists in a state of disruptive innovation and the next decade will be no different. Whether it is a shift to the cloud, hyperscale computing, software-defined data center or other technological shifts, IT must be prepared to deal with the business and pricing models that arise. Jeff Bezos is correct by not sitting on his laurels and constantly pushing the envelope in pricing and services. IT executives need to do the same and deliver comparable services at prices that appeal to the business while covering costs. This requires keeping current on technology and having the staff on board that can solve the business problems and deliver innovative solutions that enable the organization to remain competitive. RFG expects the staffing dilemma to emerge over the next few years as data centers transform to meet the next generation of business and IT needs. At that time most IT staff will not need the current skills they use but skills that allow them to work with the business, providers and others to deliver solutions built on logical platforms (rather than physical infrastructure). Only a few staff will need to know the nuts and bolts of the hardware and physical layouts. This paradigm shift in staff capabilities and skills must be anticipated if IT executives do not want to be caught behind the curve and left to struggle with catching up with demand. IT executives should be developing their next-generation IT development and operations strategies, determining skills needed and the gap, and then begin a career planning and weaning-out process so that IT will be able to provide the leadership and skills needed to support the business over the next decade of disruptive innovation. Additionally, IT executives should determine if Moonshot servers are applicable in their current or target environments, and if so, conduct a pilot when the time is right.
HP Cloud Services, Cloud Pricing and SLAs
Lead Analyst: Cal Braunstein
Hewlett-Packard Co. (HP) announced the HP Cloud Compute made generally available in Dec. 2012 while the HP Cloud Block Storage cloud entered beta at that time. HP claims its Cloud Compute has an industry leading availability service level agreement (SLA) of 99.95 percent. Amazon Inc.'s S3 and Microsoft Corp.'s Windows Azure clouds reduced their storage pricing.
Focal Points:
- HP released word that the HP Cloud Compute moved to general availability on Dec. 5, 2012 and will offer a 99.95 percent monthly SLA (a maximum of 22 minutes of downtime per month). The company extended the 50 percent discount on pricing until January. The HP Compute cloud is designed to allow businesses of all sizes to move their production workloads to the cloud. There will be three separate availability zones (AZs) per region. It supports Linux and Windows operating systems and comes in six different instance sizes, with prices starting at $0.04/hour. HP is currently supporting Fedora, Debian, CentOS, and Ubuntu Linuxes, but not Red Hat Enterprise Linux (RHEL) or SUSE Linux Enterprise Server (SLES). On the Windows side, HP is live with Windows Server 2008 SP2 and R2 while Windows Server 2012 is in the works. There are sites today on the East and West coasts of the U.S. with a European facility operational in 2013. Interestingly, HP built its cloud using ProLiant servers running OpenStack and not CloudSystem servers. Meanwhile, HP's Cloud Block Storage moved to public beta on Dec. 5, 2012; customers will not be charged until January at which time pricing will be discounted by 50 percent. Users can create custom storage volumes from 1 GB to 2 TB. HP claims high availability for this service as well and claims each storage volume automatically is replicated within the same availability zone.
- Amazon is dropping its S3 storage pricing by approximately 25 percent. The first TB/month goes from $0.125 per GB/month to $0.095 per GB/month, a 24 percent reduction. The next 49 TB prices per GB/month fall to $0.080 from $0.110 while the next 450 TB drops from $0.095 to $0.070. This brings Amazon's pricing in line with Google Inc.'s storage pricing. According to an Amazon executive S3 stores well over a trillion objects and services 800,000 requests a second. Prices have been cut 23 times since the service was launched in 2006.
- In reaction to Amazon's actions Microsoft's Windows Azure storage pricing has again been reduced by up to 28 percent to remain competitive. In March 2012 Azure lowered its storage pricing by 12 percent. Geo-redundant storage has more than 400 miles of separation between replicas and is the default storage mode.
|
Google GB/Mo |
Google Storage pricing |
Amazon S3 pricing | Amazon GB/mo | Azure storage pricing - geo-redundant |
Azure storage pricing - local-redundant |
| First TB |
$0.095 |
$0.095 |
First TB |
$0.095 |
$0.070 |
| Next 9 TB |
$0.085 |
$0.080 |
Next 49 TB |
$0.080 |
$0.065 |
| Next 90 TB |
$0.075 |
|
|
||
| Next 400 TB |
$0.070 |
|
Source: The Register
RFG POV: HP's Cloud Compute offering for production systems is most notable for its 99.95 percent monthly SLA. Most cloud SLAs are hard to understand, vague and contain a number of escape clauses for the provider. For example, Amazon's EC2 SLA guarantees 99.95 percent availability of the service within a region over a trailing 365 day period – i.e., downtime is not to exceed 250 minutes (more than four hours) over the year period. There is no greater granularity, which means one could encounter a four hour outage in a month and the vendor would still not violate the SLA. HP's appears to be stricter; however, in a NetworkWorld article, HP's SLA only applies if customers cannot access any AZs, according to Gartner analyst Lydia Leong. That means customers have to potentially architect their applications to span three or more AZs, each one imposing additional costs on the business. "Amazon's SLA gives enterprises heartburn. HP had the opportunity to do significantly better here, and hasn't. To me, it's a toss-up which SLA is worse," Leong writes. RFG spoke with HP and found its SLA is much better than portrayed in the article. The SLA, it seems, is poorly written so that Leong's interpretation is reasonable (and matches what Amazon requires). However, to obtain credit HP does not require users run their application in multiple AZs – just one, but they must minimally try to run the application in another AZ in the region if the customer's instance becomes inaccessible. The HP Cloud Compute is not a perfect match for mission-critical applications but there are a number of business-critical applications that could take advantage of the HP service. For the record, RFG notes Oracle Corp.'s cloud hosting SLAs are much worse than either Amazon's or HP's. Oracle only offers an SLA of 99.5 percent per calendar month – the equivalent of 2500 minutes or more than 40 hours of outage per month NOT including planned downtime and certain other considerations. IT executives should always scrutinize the cloud provider's SLAs and ensure they are acceptable for the service for which they will be used. In RFG's opinion Oracle's SLAs are not acceptable at all and should be renegotiated or the platform should be removed from consideration. On the cloud storage front overall prices continue to drop 10 percent or more per year. The greater price decreases are due to the rapid growth of storage (greater than 30 percent per year) and the predominance of newer storage arrays versus older ones. IT executives should be considering these prices as benchmarks and working to keep internal storage costs on a similar declining scale. This will require IT executives to retain storage arrays four years or less, and employing tiering and thin provisioning. Those IT executives that believe keeping ancient spinning iron on the data center floor to be the least cost option will be unable to remain competitive against cloud offerings, which could impair the trust relationship with business and finance executives.
Mainframe Survey – Future is Bright
Lead Analyst: Cal Braunstein
According to the 2012 BMC Software Inc. survey of mainframe users, the mainframe continues to be their platform of choice due to its superior availability, security, centralized data serving and performance capabilities. It will continue to be a critical business tool that will grow driven by the velocity, volume, and variety of applications and data.
Focal Points:
- According to 90 percent of the 1,243 survey respondents the mainframe is considered to be a long-term solution, and 50 percent of all respondents agreed it will attract new workloads. Asia-Pacific users reported the strongest outlook, as 57 percent expect to rely on the mainframe for new workloads. The top three IT priorities for respondents were keeping IT costs down, disaster recovery, and application modernization. The top priority, keeping costs down, was identified by 69 percent of those surveyed, up from 60 percent from 2011. Disaster recovery was unchanged at 34 percent while application modernization was selected by 30 percent, virtually unchanged as well. Although availability is considered a top benefit of the mainframe, 39 percent of respondents reported an unplanned outage; however, only 10 percent of organizations stated they experienced any impact from an outage. The primary causes of outages were hardware failures (31 percent), system software failure (30 percent), in-house application failure (28 percent), and change process failure (22 percent).
- 59 percent of respondents expect MIPS capacity to grow as they modernize and add applications to address business needs. The top four factors for continued investment in the mainframe were platform availability advantage (74 percent), security strengths (7o percent), superior centralized data server (68 percent), and transaction throughput requirements best suited to a mainframe (65 percent). Only 29 percent felt that the costs of migration were too high or use of alternative solutions did not have a reasonable return on investment (ROI), up from 26 percent the previous two years.
- There remains a continued concern about the shortage of skilled mainframe staff. Only about a third of respondents were very concerned about the skills issues, although at least 75 percent of those surveyed expressed some level of concern. The top methods being used to address the skills shortage are training internally (53 percent), hire experienced staff (40 percent), outsource (37 percent) and automation (29 percent). Additionally, more than half of the respondents stated the mainframe must be incorporated into the enterprise management processes. Enterprises are recognizing the growing complexity of the hybrid data center and the need for simple, cross-platform solutions.
RFG POV: Some things never change – mainframes still are predominant in certain sectors and will continue to be so over the visible horizon, and yet the staffing challenges linger. 20 years after mainframes were declared dinosaurs they remain valuable platforms and growing. In fact, mainframes can be the best choice for certain applications and data serving, as they effectively and efficiently deal with the variety, velocity, veracity, volume, and vulnerability of applications and data while reducing complexity and cost. RFG's latest study on System z as the lowest cost database server (http://lnkd.in/ajiUrY ) shows the use of the mainframe can cut the costs of IT operations around 50 percent. However, with Baby Boomers becoming eligible for retirement, there is a greater concern and need for IT executives to utilize more automated, self-learning software and implement better recruitment, training and outsourcing programs. IT executives should evaluate mainframes as the target server platform for clouds, secure data serving, and other environments where zEnterprise's heterogeneous server ecosystem can be used to share data from a single source, and optimize capacity and performance at a low-cost.