photo credit: oskay

On November 11, 2010, Pune Mirror did a story on investments related to Storage, Networking and Computing chips in Pune. The story is not available online unfortunately, but can be read in the e-Paper version, under ‘City’ section, Page 07. For those who are interested in reading this article, we are uploading a scanned copy of this story. Read it here: Pune Area Hi-Tech Investments Total $1B

Image Sourve: allSAN.com

Howard Goldstein spoke to the PuneChips community on Storage and Networking Protocols earlier this month. His presentation is now available here as a PDF file. Please download as required.

Goldstein Storage Networking – The Path to Performance

photo credit: stars6 / Leonardo Rizzi

On January 15, 2010, LSI hosted its Datacenter Evolution 3.0 forum in Pune. The Pune center is certainly getting a lot of prominence within LSI as the top two LSI execs, Abhi Talwalkar, CEO and Jeff Richardson, EVP Semiconductor Solutions Group, were hobnobbing with the attendees. VP and LSI India MD, Pravin Desale was the EmCee. Talwalkar has a strong personal connection to Pune – he was born here! He has presided over a complete makeover of the company where all internal manufacturing has been completely eliminated. LSI now is a silicon, systems and software Technology Company playing in the Storage and Networking verticals. Obviously, datacenters are a very important market.

Robinson, in his keynote, talked about how datacenters need to evolve to support the upcoming surge in data and traffic (Figure 1).

Figure 1: Surging Internet Traffic and Data; Source: LSI

His take is that new datacenters need to satisfy three key requirements – manageability, scalability and green-ness. A couple of technological innovations are powering the drive to the management and scaling of datacenters; an application aware infrastructure and storage device performance. There are also a number of innovations that reduce datacenter power consumption at all levels from device to software.

Application Aware Infrastructure

Initially, all network traffic was treated the same way, essentially as an Ethernet packet. That, however created a problem where higher priority traffic was often bottlenecked. In addition, there is no way to confirm a packet’s bona fides. The solution to this problem lies in inspecting the packet before it is forwarded to its destination. DPI or deep packet inspection technique allows looking inside a packet to identify what application it belongs to such as e-mail, VoIP, Video, HTTP, etc. and whether the packet is a virus or malware. Application awareness (Figure 2) allows infrastructure devices to meet the quality of service (QoS) requirements of the application along the entire path. With the pending move to cloud computing, application awareness is required to provide consistent performance at all points.

Figure 2: Application Aware Infrastructure; Source: LSI

An application aware infrastructure has better performance, much better levels of security and control and better management of resources. Networking giants such as Cisco Systems and Juniper Networks already use DPI in their latest generation network processors and a number of supporting devices are also hitting the market.

LSI demonstrated their Application Recognition Products at the event. It turns out that the entire development has been done by the LSI networking team in Pune.

As with any technological advance, there is a dark side; a rogue installation could use DPI to exploit an application vulnerability and to mount attacks. DPI can become a tool for govt. bodies to spy on its citizens or for organizations such as RIAA or MPAA in their overzealous attempts to fight piracy. Rumor has it that the Chinese govt. is very interested in DPI. As such, DPI vendors will need to work together with application developers to provide fool proof security to users.

Device Performance

As Ethernet speeds move from 10G to 40G/100G, an inflection point in storage device performance has been reached. SSDs or Solid State Drives offer 1000x input/output operations per second or IOPs when compared to hard disk drives or HDDs. While performance is high, SSD overhead cannot be hidden in RAID stack as you can with HDD. SSD cost is also an issue, and as such HDDs currently rule when high-capacity storage is required, but that advantage should go away once SSD volumes improve. In the current scenario, a hybrid storage that uses SSD for cache and HDD for main storage is certainly something worth looking at. The area where HDDs are expected to have a major advantage is where a large number of small files need to be stored, but that is something that could be worked around by application providers.

Here’s a recent performance comparison of SSD vs HDD performed by Tom’s Hardware.

As internet traffic and data are expected to grow by leaps and bounds in the coming years, it could be just that the next rounds of datacenter evolution (4.0, 5.0, …) may just be round the corner.