As our IT environments grow it is critical that we effectively manage and optimize our infrastructure – particularly in the storage domain where the historical doubling of capacity every 18 months has resulted in poor utilization of storage and has created islands of lost capacity.
Storage Resource Management (SRM) products aim to control this nightmare and are supposed to reduce capital costs through improved utilization as well as to reduce operating costs through automation and efficiencies.
However, deploying SRM products (such as EMC Control Center, Symantec Command Central Storage, IBM Tivoli Productivity Center, HP Storage Essentials, and others) has been littered with problems – especially in environments of any significant size and complexity. The primary reason is that SRM tools were originally architected to rely on host agents that must be installed on all servers (unix, linux, Windows) in the environment.
Why are SRM agents bad?
The first reason has little to do with technology and more to do with the way most companies have structured their IT organizations. If a storage team wants to deploy SRM agents, it needs to get permission and buy-in from the server and applications teams. Right out of the shoot, most SRM deployments are delayed by weeks or months because the server and application team wants to understand the ramification of putting an intrusive agent into their environment. What effect will it have on the performance and availability of their critical applications? How much effort and money will it cost them to support? For this reason alone, agent-based architectures are often dead on arrival.
The second reason agents fail is the high cost and complexity of deployment. Before you deploy agents across 100s or 1,000s of systems, you need a plan – and you need people to implement and manage that plan. The next time you run across a “successful” SRM deployment, try to find out how many professional services consultants where on-site and for how long – odds are that more money was spent on implementing the SRM product then was spent on buying it in the first place.
So why bother having agents in the first place? You can recognize the vast majority of storage management benefits and value without agents – in far less time and with far fewer professional services costs. NetApp SANscreen was designed and built from the beginning with an agentless architecture. This fast, flexible architecture allows SANscreen to discover large environments – even 10s of 1,000s of switch ports – within a matter of hours-days. If you could accomplish most of the value of storage management in only a matter of days without agents, why would you ever take the risk of spending months or years only to fail with an agent-based architecture?
And if very detailed host information is really required, there are plenty of industry standard technologies such as WMI (Windows) or Secure Shell (Unix) that can be used to collect host data “agentlessly.” Also, you can remotely collect data from databases like Oracle and SQL Server or even better just interface with their management tools for that data.
So if you are really interested in recognizing value from a storage management project – don’t bother with products that were originally designed to use agents. It is time to call out “Death to Agents!” – their time and their failures have come and gone.
Adios,
Paul.
