In this blog, I will cover:
- Major components of the Cloud infrastructure from hardware perspective
- 2 models of deploying Cloud infrastructure
- Overview of different converged infrastructure players and their solutions.
- Deepdive into VCE converged infrastructure solution.
Major Cloud hardware components are Compute, Storage and Network. There are 2 models of building Cloud infrastructure.
- Build using discrete components.
- Buy turnkey solution called as Converged Infrastructure.
Bigger MSDCs like Amazon, Google, Microsoft use the first model where they buy commodity hardware and connect them to create custom solution. Here, we will focus on the Converged Infrastructure model.
Converged Infrastructure model(CI):
Even though creating a datacenter with discrete compute, storage and networking options gives maximum flexibility, there are some distinct advantages in using Converged Infrastructure model.
In the CI model, vendors create a converged solution combining Compute, Storage and Networking into 1 entity and optimizing this entity to get maximum performance. Each of the components could be from same or different vendors. Since the solution is tested rather than individual components, there is no need to worry about inter-working of the individual components.
Following are some of the advantages:
- Agility – Because the solution is already packaged, time taken to turn services online is drastically reduced.
- Efficiency – Solution is optimized for performance. Maximum efficiency is derived for all components of compute, storage and networking.
- Availability – High availability aspect is taken care by CI solution.
- Ease of management – Instead of managing the disparate pieces, there is a converged management solution.
- Ease of maintenance – CI provider is responsible for any support related issues for the complete system, there is no need to approach multiple vendors.
There are multiple CI providers available today. Some of the popular ones are:
- VCE – Partnership between Vmware, Cisco, Intel and EMC. Operates as a separate entity.
- HP Cloud system.
- IBM Pureflex.
- Dell Vstart.
- Flexpod – Solution from Cisco and Netapp.
The solution from different CI providers can be compared based on:
- Integration with current datacenter hardware – Does the solution work with existing hardware?
- Openness – Does the solution allow integration with different hypervisors, Cloud OS like Openstack etc.?
- Management – Is there a good converged management solution?
- Performance and scale – Does the solution allow to scale based on the growth of the datacenter?
Following link provides a good comparison between VCE, IBM Pureflex and HP Cloud system. Instead of comparing the solutions, I will deepdive into VCE solution and provide some overview on the HP’s and IBM solution. The details below are not directed towards finding the best solution but to get a general understanding of CI components and how they are put together.
The above 2 pictures provide a good overview of components involved in IBM pureflex system. The solution has IBM servers, storage and networking switches and supports most hypervisors.
HP Cloud system uses HP servers, storage and networking switches and supports most hypervisors.
As mentioned earlier, VCE is a joint venture between Cisco and EMC with investments from Vmware and Intel. The solution uses Cisco’s UCS platform for servers, Cisco’s Nexus/Catalyst switches for networking, VNX and VMAX storage systems from EMC and Virtualization software from Vmware. All these entities are packaged and sold as a solution.
Vblock is the name of the combined product and based on the datacenter scale needed, there are different Vblock systems like Vblock 100, 200, 300, 700.
The above picture shows the different compute, network, storage and virtualization pieces of Vblock 100.
The above picture shows the topology of compute, storage and networking pieces.
The above picture shows performance numbers with Vblock 100.
- Number of servers can vary from 3-8.
- Storage capacity can go upto 52TB.
Vblock 700 is the highest end product available from VCE and combines upto 384 servers.
The above picture shows the different hardware components associated with Vblock 700. Comparing with Vblock 100, instead of rack servers, blade servers are used and instead of Catalyst switches, Nexus switches are used.
The above picture shows topology diagram of Vblock 700.
When a single Vblock system gets completely used up, the enterprise or provider can buy 1 more Vblock system and integrate the 2 into a system. The picture below illustrates how to do this.