Just when you’ve mastered the distinctions between the different flavours of “Cloud”, along comes the Converged vs Hyperconverged challenge. If you are confused you are not alone. Stack Group’s Chairman Jeff Orr sets out to draw the distinction and examine the use cases.
The differences between Convergence and Hyperconvergence
Jeff says “Converged is called the second phase of the convergence process while “Hyperconverged” claims the third phase title suggesting it’s an evolution and improvement of the second. In fact they are different and are going to co-exist as solutions for different uses”.
The first phase of the convergence process began with virtualisation and was rapidly adopted because of its many and obvious benefits. Unfortunately it brought with it some new problems. In particular, storage management became complex and difficult to scale while application and storage tiers created silos of data.
Virtualisation - The continuous increase in data volumes has led to storage management issues
By combining the separate hardware element of a virtualised infrastructure into one standardised solution many issues can be reduced. This is the second phase of convergence. Standardisation simplifies deployment, support and even procurement. Manufacturers put together “Gold Standard” templates of storage, compute and switching which are pre-tested and performance validated. This instils confidence in procurement, simplifies deployment and support and shortens the sales cycle by reducing the need for proof of concept. Cisco’s UCS, EMC’s Vblock and NetApp’s FlexPod are examples of this phase of convergence.
In the third phase, “Hyperconvergence” the notion of combining these separate components is dismissed. Instead software is used to enrol multiple nodes of conventional commodity servers into a resilient cluster that provides the functions of storage, compute and switching. Some manufacturers provide this software separately, others provide it as an appliance with it installed.
Hyperconvergence solutions scale easily by adding nodes to the cluster
Software vendor VMware partners with hardware vendors in its EVO:RAIL program whilst SimpliVity, NutaniX and Scale computing are examples of appliance vendors.
Hyperconvergence solutions scale easily by adding nodes to the cluster or by federating clusters. Resilience is provided by node redundancy and as there is no separate storage array (it’s all direct attached within the servers) management is simplified. The virtual machines manage the storage that is local to the node on which they reside. Indeed the portmanteau word Hyperconvergence is said to come from the hypervisor of virtualisation and convergence.
Hyperconvergence suits deployment of Virtual Desktops and primary storage for SME’s and small multi-branch organisation’s remote offices. Because it needs to write a second copy of every storage block to another node in the cluster in case a node fails capacity efficiency is low. De-duplication is needed to overcome this but ties up CPU cycles and IOPS.
Convergence at phase two is still best option for economically providing for heavy transactional workloads needing high IOPS such as databases, R&D and big data.
Written by: Jeff Orr, Chairman of the Stack Group
Jeff Orr founded the Stack Group in 1984, and prior to this owned and ran several other successful IT businesses. He is a qualified chartered engineer and winner of the Bibby Prize for Engineering from the University of Liverpool.
Jeff Orr is highly respected in the industry and consults for some of the biggest football clubs in the world.
About Stack Group
The Stack Group deliver solutions and support in the areas of IT Management and Cloud services, Telecoms, Infrastructure and Networks. If you would like any help and advice regarding, Cloud services, data services, infrastructure or telecoms please do not hesitate to get in touch and one of our experts will assist you.
Phone: +44 (0) 151 521 2202
Email: [email protected]
The differences between Convergence and Hyperconvergence
Jeff says “Converged is called the second phase of the convergence process while “Hyperconverged” claims the third phase title suggesting it’s an evolution and improvement of the second. In fact they are different and are going to co-exist as solutions for different uses”.
The first phase of the convergence process began with virtualisation and was rapidly adopted because of its many and obvious benefits. Unfortunately it brought with it some new problems. In particular, storage management became complex and difficult to scale while application and storage tiers created silos of data.
Virtualisation - The continuous increase in data volumes has led to storage management issues
By combining the separate hardware element of a virtualised infrastructure into one standardised solution many issues can be reduced. This is the second phase of convergence. Standardisation simplifies deployment, support and even procurement. Manufacturers put together “Gold Standard” templates of storage, compute and switching which are pre-tested and performance validated. This instils confidence in procurement, simplifies deployment and support and shortens the sales cycle by reducing the need for proof of concept. Cisco’s UCS, EMC’s Vblock and NetApp’s FlexPod are examples of this phase of convergence.
In the third phase, “Hyperconvergence” the notion of combining these separate components is dismissed. Instead software is used to enrol multiple nodes of conventional commodity servers into a resilient cluster that provides the functions of storage, compute and switching. Some manufacturers provide this software separately, others provide it as an appliance with it installed.
Hyperconvergence solutions scale easily by adding nodes to the cluster
Software vendor VMware partners with hardware vendors in its EVO:RAIL program whilst SimpliVity, NutaniX and Scale computing are examples of appliance vendors.
Hyperconvergence solutions scale easily by adding nodes to the cluster or by federating clusters. Resilience is provided by node redundancy and as there is no separate storage array (it’s all direct attached within the servers) management is simplified. The virtual machines manage the storage that is local to the node on which they reside. Indeed the portmanteau word Hyperconvergence is said to come from the hypervisor of virtualisation and convergence.
Hyperconvergence suits deployment of Virtual Desktops and primary storage for SME’s and small multi-branch organisation’s remote offices. Because it needs to write a second copy of every storage block to another node in the cluster in case a node fails capacity efficiency is low. De-duplication is needed to overcome this but ties up CPU cycles and IOPS.
Convergence at phase two is still best option for economically providing for heavy transactional workloads needing high IOPS such as databases, R&D and big data.
Written by: Jeff Orr, Chairman of the Stack Group
Jeff Orr founded the Stack Group in 1984, and prior to this owned and ran several other successful IT businesses. He is a qualified chartered engineer and winner of the Bibby Prize for Engineering from the University of Liverpool.
Jeff Orr is highly respected in the industry and consults for some of the biggest football clubs in the world.
About Stack Group
The Stack Group deliver solutions and support in the areas of IT Management and Cloud services, Telecoms, Infrastructure and Networks. If you would like any help and advice regarding, Cloud services, data services, infrastructure or telecoms please do not hesitate to get in touch and one of our experts will assist you.
Phone: +44 (0) 151 521 2202
Email: [email protected]