Project Log

Moving ahead with IP-SAN

MphasiS wanted to consolidate its storage, so the company deployed an IP-SAN solution. Aditya Menon, the company’s group chief information officer, shares insights about this project

Prior to going in for an IP-SAN deployment, we had discrete individual storage spread across our organisation. There was no storage consolidation and consequently no uniformity of storage. We had more than 1,000 servers running across locations, each with its own storage set-up. Managing

individual storage in diverse locations was a nightmare as we had to take a tape backup for each individual storage. This was time-consuming, and took, on an average, eight hours for one TB of data per location. Our disaster recovery (DR) process consisted of sending tapes to a far-off location, but then getting the tape online used to take, on an average, more than 24 hours for one TB of data at each location. Finally, restoring that data was also time-consuming, and it took around 48 hours to recover an average 200 GB. Replication and disaster recovery were therefore becoming a big challenge. As far as backups were concerned, our locations were taking full backups on a daily basis.

Going for an IP-SAN

We wanted to do away with the pain points in storage and smoothen replication, hasten backup, and ensure the efficient management of our storage infrastructure. We also wanted to consolidate our storage for voice recording and mail servers. The key driver for investing in IP-SAN technology was to lower the TCO (total cost of ownership). The IP-SAN was required for file and e-mail servers. With the coming of Gigabit Ethernet we have faster access to file servers and can plug in all our servers to the IP-SAN ports. This makes IP-SAN much faster and data retrieval is almost comparable to that of an FC-SAN.

Zeroing in on Intransa

After a stringent evaluation exercise we chose an IP-SAN solution called IP-5500 from Intransa Inc, having found it to be the best from a storage management perspective, especially in backup management. For instance, in case I lose a file server, I can connect to the IP-SAN box and bring it back without looking for another server. We evaluated many vendors and found Intransa to be the best bet. The evaluation exercise started in end-2004 and involved many sessions. The implementation work has just started; we intend to install five boxes and may scale up later. We have a team of eight members involved in this exercise, and we expect our first box to go live by the first week of March 2005.*

Great expectations

Although we are still in the process of implementing the IP-SAN solution, I have great expectations from it and hope that it will be the best fit for our requirements. We feel that an IP-SAN topology offers an application server that is independent of storage and data services. In case the application server goes down, the data is still available on the IP-SAN, and hence another server can quickly access that data and provide it to users. Thus, with an IP-SAN, there is no chance of a service outage due to the application server going down. Additionally, IP-SAN technology provides for the rapid backup and restoration of data without using LAN resources. That said, it facilitates LAN-free backup and very fast restore operations. We believe that restoration time will come down from 48 hours to less than an hour. We expect to save at least $20,000 during phase one of the IP-SAN implementation.

An IP-SAN also provides SAN-based replication—and thereby DR-ready data management—at a reduced TCO. The IP-SAN eliminates guesswork with regard to expansion by providing scalability on a pay-as-you-need basis. It also reduces the cost of ongoing server expansion as it is easy to scale servers using this technology. The best part of using IP-SAN technology is that it simplifies storage management.

A welcome change

By going in for networked IP-SAN storage, we have been able to bring in uniformity in our storage resources spread across the globe. This will give us a modular storage architecture. Online replication will ensure that our users are able to access replicated data in real-time. As storage will now be consolidated, we will be able to manage storage resources better. There will no longer be individual storage attached to different servers. Smooth DR operations will be facilitated by storage consolidation. The earlier process of taking a tape backup, sending tapes to a far-off location, and then uploading that data online will not be required as everything will be online.

As told to Abhinav Singh
* This article was written in February—Editor