US-based research firm Ashton, Metzler & Associates has found that many SANs just stop causing applications to slow down or suffer what the firm calls a 'brownout'. The firm surveyed just over a 100 IT professionals in Global 2000 and other companies, and found that more than 80 per cent of the companies surveyed ran a SAN. A quarter have more than five SANs each. Seventy per cent of the people surveyed said that SAN slowdown affects their applications once a month. A quarter of the people surveyed even said they suffered a SAN slowdown once a week.
If these figures apply to all organisations running a SAN then almost 90 per cent of them have a SAN outage twelve times a year. The results are throttled back applications, loss of efficiecy and management time expended. Backups could take longer than expected or there might be extra help desk calls. There could also be revenue loss if an ERP or similar application was affected by a SAN slowdown.
The researchers were looking at SAN adoption and the experiences of organisations with SANs.
SAN adoption has become mainstream, but they found that, "approximately half of respondents (indicated) that they do not have a good handle on what causes brownouts." A supposition is that a large number of server ports are trying to access a smaller number of storage ports and that the storage network might not have sufficient bandwidth. Respondents also pointed out that they didn't have access to good SAN performance reporting and trending tools.
Neither did they have access to good overall fault location tools. Faults could occur in the application server, the storage network path with its HBAs, ports, switches, inter-switch links and possible core director switches, the storage network software and storage arrays. SAN fault location is in its infancy.
If these results apply generally to all SAN-operating organisations then we might suppose that the overall storage networking infrastructure is becoming much more complicated with storage networking hardware and software layers multiply.
Perhaps what Techworld has speculated about application infrastructures applies to storage networking also; "We've reached the stage now where management of (storage networking) is becoming more and more complex, perhaps to the point where it is beyond the capability of the average IT administrator to understand what is happening at any specific point in time. We're building multi-tiered server and storage infrastructures the bits of which we understand but the totality of which is beginning to exceed our grasp."
A recent contribution to Techworld praised the flexibility of direct-attached storage. It is also apparent that DAS is simple whereas SANs are not, especially SANs with virtualisation and complex fabrics.
On the other hand only a few organisations were surveyed, just over 100 out of the tens of thousands world-wide using some form of storage networking. Whether the problems reported are limited to the few or visited on the many though, SAN management to detect, repair and prevent SAN slowdown is vital. If it doesn't exist then the value of a SAN is far less than it appears at first hand.