Cloud architectures are getting complex, partly because of the range of services that they offer, and partly because of the rapid innovation that is designed to reduce the total cost of ownership (TCO) of the cloud infrastructure. You probably heard about HyperConverged infrastructure and distributed storage solutions that are now gaining rapid adoption. In addition to datacenter hardware, software is also evolving rapidly due to the availability of Opensource cloud systems like OpenStack and KVM. Various cloud architectures have varying levels of performance and scale.
The performance and scale of clouds are very important if you are a Cloud or a Service Provider because you are selling this as a service. At the time of writing this, there is a woeful lack of guarantees by providers. A key reason behind the lack of strong SLAs is that there is no way to test the entire cloud infrastructure as one. We solve this very problem with the new Spirent HyperScale solution, the ultimate scale test of a cloud.
Filling in a big gap in Cloud testing
Ultimate scale test
Ask any cloud user, and he/she would swear by sudden intermittent performance issues. This could be because of over-subscription of resources resulting in unrealistic multitenant or starvation of services. The reason these conditions are not detected during design time is the lack of full scale testing. Spirent’s HyperScale solution can help you test using as many as million virtual machines (VMs).
End-to-end cloud testing
There are several gaps in testing a cloud data center today. Say you are thinking about an overlay or underlay network design for your cloud; maybe with a SDN controller. With tradition test tools, you can test the underlay and overlay networking components and the SDN controller separately and with the fabric. However, in order to see what the end user sees, you need to measure the characteristics from the VM perspective. Spirent HyperScale helps you do this, because this is an end-to-end test solution that measures at the VM level.
Cloud QoS testing
Cloud networking software and storage software now implement fine grain QoS policies. This could be exposed to the cloud customers through varying tiers of usage. However, these technologies are not leveraged well for cloud users because they cannot be tested. Now, using Spirent HyperScale, you could configure 10,000 VMs with higher QoS and the rest of the VMs with best effort, and test this with a Million VMs scale. Spirent HyperScale receives measures from every single VM in real time; which you can detect even if a single VM sees something odd in the entire cloud.
Ensure your Cloud design meets your performance requirements
To evaluate your cloud architecture with different components from different vendors put together all the pieces – hardware, software, hypervisor, SDN etc. and run Spirent HyperScale. The solution will connect with your cloud, and populate as many VMs as your cloud can hold. Then run a test case that you have designed, and measure the performance of your cloud. Change your cloud design and repeat.
You can use these results to find if your cloud design meets your performance requirements. You can also use to avoid unnecessary over-provisioning but still meets your business requirements – this could save millions of dollars in infrastructure costs.
Spirent HyperScale test solution connects with your cloud, spins many VMs, and runs testcases using the entire cloud.
Spirent HyperScale first spins up the VMs in your cloud. It is possible to clone these VMs in any way the user requires them to be. Each VM can be configured with specific vCPU, memory, network and storage. You could describe a virtual topology with a million VMs and Spirent HyperScale can create this for you. This topology could be a pairs, full-mesh, multi-tenant etc. The VMs it starts are Spirent TestCenter Virtual VMs are specially instrumented which are capable of running any kind of data traffic or storageIO. STCv is a completely virtual product that is available from Spirent separately. Spirent HyperScale uses STCv at very large scale.
Then Spirent HyperScale then connects with these VMs to synchronize. It is very important to synchronize the VMs to coordinate the test case. If it is not synchronized some VMs could go ahead of others and the user would not be able to control the test accurately.
Once the VMs are synchronized, the user is given control of the test. Remember the entire cloud is under test, as we would call it “Cloud under Test” (CUT). With a push of a button the user can light up an entire cloud instantaneously, which will engage all the VMs across the different tenants, the networking and the storageIO.
The VMs are designed to report live results which are funneled to the web GUI. The user could see live test results from a million VMs. A single packet drop or a storageIO failure across the entire cloud can be easily viewed.
Define your service tiers and analyze their cost to you
If you are a Cloud or Service Provider, you must ensure “Money Back” SLAs. Cloud customers demand it. Spirent HyperScale can help evaluate if the infrastructure (software and hardware) can live up to the scale and performance requirements of the SLAs. Spirent HyperScale could also be helpful in defining your tiers and analyzing the cost of the tiers.
If you are a cloud solution architect or a cloud software vendor, you need to prove the scale and performance numbers of your offering. You can use Spirent HyperScale to test your solution, and advertise your impressive scale numbers.
Learn more about Spirent HyperScale test solution and how it can help you better guarantee SLA’s to your customers check out the following resources:
Announcement – Industry’s first Spirent HyperScale test solution