Managing complex data sets with PolyServe Matrix Server
Shared by: hnr19912
Managing complex data sets with PolyServe Matrix Server The amount of data managed by IT departments is growing at an ever-increasing rate – as are the potential problems it creates. Managing access, dealing with hardware failures and increased workloads are a growing challenge for IT professionals everywhere. Fortunately, there is help available, in the form of the PolyServe Matrix Server. This neat device simplifies problems and adds a much-needed layer of reliability. In today’s business world, databases are literally everywhere. They underpin mail servers such as Microsoft Exchange, hold documents in collaboration solutions such as SharePoint and Domino and are at the heart of document management products. Business Intelligence is all about mining corporate data to find a competitive edge. And the more databases we have, the more users there are that need to access and query the data. As the amount of data has grown, the logical and computation part of databases has – out of necessity – been separated from the physical data itself. In many datacentres today, the physical data is located on the Storage Area Network while the actual processing of the data is done by application and database servers. The larger quantities of data has also meant greater numbers of actual servers – while work teams are increasingly spread out across continents and even the world. But of course, to provide the most efficient response, servers and work teams need to be located as close together as possible. When a workload needs to be moved or when a server fails, users need to migrate their work to another server, as seamlessly, and quickly, as possible. Should the network or the backend storage systems fail, then data requests need to be directed to other locations where there are copies of the data. All this is done through clustering and workload balancing. But here’s the real problem. Even the slightest problem for one data management application can have a serious knock-on affect that can cause data to become unavailable to other applications. This can cause a cascade failure of applications and, put simply, spell big and expensive problems for businesses of all types. This is where PolyServe Matrix Server comes in. The PolyServe Matrix Server (PMS) uses a number of technologies to ensure high availability of data and high throughput of workload without requiring large numbers of operators or excessively complex systems. To abstract the physical location of the data from the users, PMS uses a virtual device layer (VDL) that hides where the data is. This VDL does several things. Firstly it ensures that PMS always sees the underlying data location as the same name. Even if the real location has to be changed because of a network or hardware failure, PMS and the applications that sit on top of it never know. VDL also acts as a gatekeeper to prevent any application trying to carry out an action that would impact on other applications sharing the data storage. An example of this is preventing a device from being formatted when it is being used by other applications. To ensure data consistency, PMS uses a Symmetric Distributed Lock Manager (DLM). Unlike a lot of distributed data solutions that rely on a single master locking server that can become a serious bottleneck, DLM is deployed on all application servers that will access the data. Applications lock data and records in order to update them or when they add new information. DLM uses a proprietary locking protocol so that all the servers know what is locked and use that to minimise delays in writing data. Such locking solutions are not uncommon but while other vendors broadcast information about locks indiscriminately across the network causing massive increases in traffic, the DLM locking protocol targets just the servers that are sharing the data. To speed up the delivery of data to the user or application, data is often moved from the database to a local cache. This means that data is available at the speed of the system not the network. The problem here is that caches can quickly get out of date and users can be working on old data sets. By using a Cache Coherency mechanism, PMS allows the different cache pools to know what has been updated so that they can always serve the user or application the latest set of data. To keep all of this together across multiple databases, storage devices, users, servers and physical locations, PMS has a single management interface. This reduces operator time and ensures that the corporate data assets can be easily managed. It also adds another level of high availability as the management can be done from any site. Keeping track of, and managing access to, large amounts of complex data is critical to keeping a business running. PMS provides a software layer that brings together the hardware and the applications to provide users with the highest availability of data.