Converged storage/compute stacks have become increasingly popular over recent years, with all major IT suppliers putting converged offerings into their portfolios. IDC estimates the total converged storage/compute infrastructure market will be worth $17.8bn by 2016.
Why converged storage/compute stacks?
Today’s IT departments have a bewildering choice of technologies as they build and maintain their datacentres. To cope with this, larger organisations have built teams around the technology disciplines of server, storage and networking and the best-of-breed solutions in each area.
More on converged stacks
The luxury of dedicated teams isn’t available to all organisations, however. Many IT departments have to be multi-disciplined and deal with strategic and operational tasks.
Converged stacks can help reduce the overall cost of deploying and managing IT infrastructure and provide an alternative for companies that can’t or don’t want to invest time and effort into integrating bespoke solutions. Suppliers that offer converged infrastructure deliver added value to the customer by providing validated configurations based on tested sets of product components.
There are two variations of converged storage/server stacks available today. There are those sold as a bundle of components that includes compute, networking and storage. Then there are reference architectures that describe configurations of server/storage/networking products validated by the vendor but which must be procured separately.
Converged stacks benefits and drawbacks
Bespoke solutions require a significant amount of effort to design and maintain. Anyone involved with IT upgrade or transformation projects will recognise the issues around maintaining interoperability matrices, keeping systems up-to-date and validating firmware, driver and patching configurations. These problems increase as environments scale.
Converged stacks can simplify the issues of managing infrastructure. Some of the benefits include:
- Quick deployment – Systems can be designed, built, configured and deployed in a matter of days or weeks because much of the work has been done by the integrator.
- Single support point – Also known as having “one throat to choke”. The support model is simplified by having one supplier to address all issues. This removes the possibility for disputes between vendors when complex problems arise.
- Certified solution – The vendor does all the certification and testing work up front, validating which configurations will work with each other.
- On-going maintenance support – Suppliers typically provide tested and validated maintenance rollups as software patches, code releases and new software versions are released for the individual components.
Of course, converged storage/server stacks are not a panacea for all problems. There are some disadvantages to using converged stacks, including:
- Fixed configurations – Some offerings can be difficult or impossible to modify from the purchased design.
- Lack of component choice – Suppliers will bundle only their own products, making choice limited and potentially excluding best-of-breed components. This can also include limits on available software such as hypervisors.
- Cost – Converged solutions aren’t always cheaper than do-it-yourself solutions on purely hardware alone, and there can be a premium for the added value. However, the TCO of deploying infrastructure should be the guide rather than simply hardware cost.
As converged offerings have matured, we have seen the emergence of so-called hyper-converged infrastructure. These solutions take things a step further than just bundling components into a single rack. Hyper-converged platforms combine the functions of compute and storage into a single set of hardware (usually commodity servers) that provide compute and storage functionality. Solutions in the market are available from Nutanix, SimpliVity and Pivot3. These are outside the scope of this article, however.
Converged stack supplier roundup
Converged infrastructure offerings are available from EMC and Cisco’s VCE, NetApp and Cisco’s Flexpods, EMC (VSPEX), Oracle, IBM, Hitachi, HP and Dell. Some of these are product bundles, while Flexpod and VSPEX are reference architectures.
Most vendors (with the exception of VCE) offer support for a range of virtualisation hypervisors. All suppliers provide solutions for infrastructure-as-a-service and applications, including traditional databases and big data.
VCE is a coalition between EMC and Cisco Systems plus VMware. The three companies bring their products together to create Vblock converged stacks (pictured), with EMC providing the storage, Cisco providing networking and servers and VMware supplying the hypervisor software layer.
The coalition sells pre-built and tested stacks in fixed configurations. At the top end, the Vblock 700 uses Cisco UCS blade servers, Cisco MDS and Nexus switches for the network components and EMC VMAX arrays for storage. Midrange 200 and 300 Vblock series use EMC VNX storage, while the Vblock 100 uses the EMC VNXe. Within each hardware layer, the quantities and specific configurations can be varied to customer requirements.
Management of Vblock solutions is achieved with a number of existing software tools, including Cisco UCS Manager, EMC Unisphere and VMware’s vCenter Server. This can also be supplemented with VCE Vision, which provides additional functionality including external management through an open API.
There are also VCE specialised systems for high performance databases and SAP HANA. Finally, there are some Vblock solutions for data protection that re-bundle EMC’s existing backup product range.
Flexpod is a reference architecture that uses technology from NetApp and Cisco Systems. Architecture designs include Cisco UCS servers, Cisco Nexus switches and NetApp FAS and V-Series storage virtualisation appliances. Virtualisation support includes VMware and Hyper-V.
Although Flexpod solutions are flexible in design, they have some sensible minimum requirements, including dual switches and servers for redundancy (as well as their interconnects) and NetApp FAS controllers in high availability (HA) pair configurations. There are similar minimum requirements for software.
Management of Flexpod solutions is provided through the use of NetApp OnCommand and Cisco UCS Manager. Full orchestration features are provided through Cisco UCS Director, including the ability to manage solutions using an open API.
Flexpod offerings come in three flavours. Flexpod Datacentre (including VMware, SAP, Microsoft and Oracle products) is designed for large environments. FlexPod Express is targeted at small and medium-sized businesses. Finally, there is FlexPod Select, which is targeted at high performance workloads, including Hadoop big data analytics.
As well as being part of VCE, EMC provides a number of reference architectures under the VSPEX brand. These designs are very flexible but in general are based on EMC storage products including VNX and VNXe. Servers are either Cisco UCS or Intel rackmount x86. Networking is Cisco or Brocade, and hypervisors can be VMware, Microsoft Hyper-V or Citrix XenServer.
The bundles offered by EMC are:
- VSPEX End User Computing
- VSPEX for Virtualised Microsoft Exchange
- VSPEX for Virtualised Microsoft SharePoint
- VSPEX for Virtualised Microsoft SQL Server
- VSPEX for Virtualised Oracle
- VSPEX for Private Cloud
Oracle offers converged storage/server products with a specific application focus under what it terms “engineered systems”. This includes the Exadata Database Machine, the Big Data Appliance, Exalogic Elastic Cloud and Oracle SuperCluster. All these products take components from Oracle’s portfolio, including their server technology and storage products to deliver unified solutions focused on providing easier access to Oracle’s database products.
IBM offer solutions divided into three areas, all marketed under the PureSystems brand. PureFlex System provides general infrastructure solutions to meet infrastructure-as-a-service requirements. Meanwhile, PureApplication offerings integrate IBM database and middleware offerings, including DB2 and WebSphere. Finally, PureData is a set of integrated appliances that deliver analytics features, including Hadoop.
PureSystems offerings use IBM StoreWize v7000 series storage, IBM networking products and IBM server blade chassis products that support both Intel x86 and IBM Power Systems. Virtualisation software is comprehensive and covers vSphere, Hyper-V and Red Hat KVM. There are options to use other IBM storage products in some high-end configurations.
IBM provides Flex System Manager, deployed as a dedicated appliance to manage PureFlex systems. This uses existing IBM software tools, including Tivoli Storage Manager.
Hitachi Data Systems
Hitachi Data Systems (HDS) uses the Unified Compute Platform (UCP) as the branding for its converged infrastructure offerings. These are split into pre-integrated solutions and reference architectures.
The pre-integrated platforms combine either HDS’s own compute blades (normally only sold separately in Japan) or Cisco UCS servers with either Brocade or Cisco switching and Hitachi storage. Storage options include VSP, HUS-VM and HUS platforms depending on the scale and performance of storage required.
At the hypervisor and application layer, HDS provides solutions for SAP HANA, Oracle (both single instance and Oracle RAC clusters),VMware vSphere and Hyper-V.
The management of HDS offerings is brought together with UCP Director. This software provides a single management framework for all components of the infrastructure, with full vCenter integration, federated management of multiple UCP environments, automated non-disruptive firmware upgrades. All of this can be integrated and driven from the customer’s own management tools using a REST-based API.
HP has been offering converged systems solutions for some time. Its products use HP blade systems, HP 3PAR storage and HP’s own networking technology. Solutions include ConvergedSystem for Virtualisation, ConvergedSystem for Cloud, ConvergedSystem for Big Data, ConvergedSystem for Client Virtualisation and ConvergedSystem for Collaboration.
HP has recently released OneView, a storage orchestration platform that provides deeper integration into the individual components of the ConvergedSystem stacks. This enables tasks such as HP firmware upgrades to be managed from a single interface rather than separate element manager tools.
Dell vStart bundles include Dell servers, storage and switches, and come with VMware or Hyper-V software. The vStart series come in two configurations – “v” for VMware equipped and “m” for Hyper-V. The vStart 50 is the smallest configuration and it’s designed to support 50 virtual machines (VMs). Bundles then go through the vStart 100, vStart 200 and up to the vStart 1000. vStart configurations use PowerEdge servers, EqualLogic or Compellent storage and Dell switches.