ProfitBRICKS IAAS Live Vertical Scaling Add more data center infrastructure resources on demand, without a reboot Reconfiguring during ongoing operation a world debut. Using ProfitBricks Live Vertical Scaling, you can increase the number of your Cores as well as your RAM without having to restart your virtual server. Expanding your virtual data center during peak loads is therefore possible while your servers are running. ProfitBricks Live Vertical Scaling: absolutely unprecedented increasing resources in ongoing operation!
Live Vertical Scaling: Comfortable and Cost-efficient Cloud computing solutions have long promised the ability to create servers and networks instantly, adding capacity when needed. It s well established that Cloud computing IaaS solutions are far more cost effective and efficient, than running traditional physical data centers, however only those solutions designed to scale vertically, as well as the more common horizontally scaling approach, deliver on the promise of cloud computing. With vertical scaling (scale up), system performance is increased by adding resources (CPU cores, RAM, storage space, etc.). Imagine a data center where physical resources can be added to an environment without requiring modifications to the software, databases or middleware. This is ideal, for example, for LAMP (Linux, Apache, MySQL, PHP) environments, as MySQL recognizes the new resources, without any modifications, and instantly benefits from the additional resources without a restart. Most other cloud computing services are designed only for horizontal scaling (scale out), in which the total performance is increased by adding additional instances or virtual serv- ers. To maximize applications for horizontally scaled infrastructure, software architects and developers need to parallelize the software or modify configurations. For example, scaling Ruby on Rails applications to scale-out infrastructures can be very complex from a technical point-of-view and thus, expensive. Additionally, horizontally-scaled server environments are inevitably more complex and thus more difficult to administer. Troubleshooting is also far more complicated should a problem occur. Furthermore, the performance of horizontally scaled environments does not increase on a completely linear basis with the number of servers/nodes, as communication between the servers entails an administrative overhead. Using several operating systems also involves overhead. These additional costs are what eat up a part of the maximum theoretic output. A number of advantages in every day life The value of vertical scaling can be best illustrated using practical examples. ProfitBricks IaaS Live Vertical Scaling Page 2
Flexibility in Development and Testing For instance, software developers can benefit from massive increases in CPU cores and RAM on request. For CPU-intensive tasks such as compiling larger-scale projects, developers simply allocate more CPU performance via a script and the API from the provider of the Infrastructure as a Service (IaaS) offering. Development teams with access to a vertically scaling IaaS provider can be more agile and have infrastructures that resemble the subsequent production environment. They can test more efficiently, for example, and could ask - Should an entire database be retained in the memory? Or they could run additional testing for example a load test simulating several thousand users starting the application? In a vertically scaled, virtual environment, this is not a problem. These types of tests reveal the code quality and if problems occur in testing, inadequate hardware performance can most probably be ruled out. If any difficulties do occur, it is most probably due to the pre-released code. Fig. 1: Vertical Upscaling im DCD ProfitBricks IaaS Live Vertical Scaling Page 3
Scale Up via Slider IaaS Providers like ProfitBricks offer users the opportunity to allocate the required CPU cores and memory resources on a granular basis. RAM, for example, can be additionally allocated or de-allocated in steps of 1 gigabyte. Unlike traditional VPS providers, ProfitBricks does not have fixed plans and packages with specific amounts of CPU cores/ram/storage. For each virtual server, up to 48 physical CPU cores and 196 Gbyte or RAM can be booked in the ProfitBricks data center. Share resources, reduce costs Granular, API-driven resource allocation and de-allocation can also be an ideal solution for IT infrastructure, such as collaboration system or file servers in medium-sized companies, which typically only work in one time-zone and therefore have a low usage period overnight. In these cases, any CPU cores and RAM that are not required would be de- allocated in the servers on which the collaboration system is running via a script, until the start of the heavy load phase in the morning. Full-capacity is once again available punctually at the start of the working day. With ProfitBricks, minutebased billing customers could experience great cost savings In this type of environment. With ProfitBricks Live Vertical Scaling and a graphic configuration tool for creating and configuring virtual data centers (called the Data Center Designer or DCD), CPU cores and RAM resources can be additionally allocated and de-allocated via a slider button. For instance, the DCD will increase processor cores by signaling to the operating system via an ACPI (Advanced Configuration and Power Interface) event. If the system understands this signal, it can use the new resources without a restart. With other IaaS providers, users need to first shut down their servers, then allocate the new resources and finally restart the systems. Even in a best-case scenario, the servers are still down for several minutes, so that this type of modification, without Live Vertical Scaling, is only possible during off hours. No reset needed ProfitBricks provides its customers with reference operating system including Debian and ProfitBricks IaaS Live Vertical Scaling Page 4
Reliability and High Availability Ubuntu Linux distributions free of charge, both of which include this functionality. Other Linux variants with current kernel versions are also proficient in handling a dynamically changing number of cores. It can be assumed that future versions of other server operating systems will also be able to recognize CPU cores without a restart natively and without any need for modifications. As the KVM (Kernel-based Virtual Machine) hypervisor used by ProfitBricks can trigger the ACPI event required for Live Vertical Scaling, ProfitBricks has modified parts of the hypervisor accordingly. The company provides these and other modifications such as new kernel modules as open source to all KVM users. Data security, guaranteed The virtualization of servers at IaaS providers regularly prompts customers to ensure that the data they have stored in the cloud is protected. ProfitBricks ensures, from the technical point of view, that no inadvertent, unauthorized data access by other customers occurs every user is allocated physical CPU cores exclusively, even if this leads to idle times on the part of the provider. With ProfitBricks IaaS, customer CPU cores are not shared among other customers. The same applies for RAM, the memory address ranges are assigned exclusively to the respective virtual machines (VM). Bad neighbor situations, as they are known, are never a problem for ProfitBricks customers. In addition to the higher level of data security, typically not included as a matter of course from other IaaS providers, permanent assignment also affords a further benefit: customers can rely on the performance of their systems. ProfitBricks guarantees that the resources that have been allocated are permanently available. If the physical server on which a particular customer VM is running has insufficient resources to handle the values required via DCD of API requests, the hypervisor used by ProfitBricks will automatically move the virtual machine to another server. This all happens seamlessly in the background and as no negative impact on the applications running in the VM. The RAM image is simply transferred to the new server ProfitBricks IaaS Live Vertical Scaling Page 5
Practice-oriented and User-friendly via ProfitBricks s exclusive InfiniBand highperformance infrastructure (maximum data throughput: 80 gigabit per second) and the connections to the respective virtual storage systems are retained. The advantages described above clearly show that cloud computing services with vertical scaling is the ideal Infrastructure as a Service architecture. ProfitBricks is the only IaaS provider to offer Live Vertical Scaling, plus customers benefit from dedicated CPU cores, and RAM and the easy configuration of these resources via the Data Center Designer or API. Copyright 2012 ProfitBricks GmbH www.profitbricks.com Greifswalder Str. 207 10405 Berlin, Germany email: lvs@profitbricks.com Hotline: 0800-22 44 668 INTERNATIONAL CONTACT ProfitBricks Inc. 15 Tudor Street Cambridge, MA 02139 email: info-us@profitbricks.com Hotline: +1-866-852-5229 ProfitBricks IaaS Live Vertical Scaling Page 6