We ran an RSS version of Hu Yoshida's blog for several years on this page, and it is excellent reading. If you are looking for it, you can find a link in the menu below.
This site provides hints and tips on how to manage your data, strategic advice and news items. A lot of this advice is about smarter, cheaper ways of working without compromising service.
Click on the links above to open up a side menu for the different sections of the site, or just select from the highlights further down the screen.
This time last year, we were hearing rumblings about a new virus in China, but no-one could predict how Covid-19 would hold the business world hostage in 2020. We all learned new words like lockdown, pandemic and self-isolation. We took part in online meetings as we were told to work from home. Many who could not work, were paid by the governments to stay at home. Superhuman efforts were made to find a vaccine, and vaccines were made and approved in record time.
The Black Lives Matter protests spread throughout the world and changed attitudes. Then there was the small matters of an election in the US, and the endless Brexit deadlines in Europe.
While we are still firmly in the grip of Covid-19, with the virus spreading rapidly again in the Northern Hemisphere as winter begins, there is hope for 2021. As more vaccines are produced and administered, the virus should loosen its impact. Politics should settle down with the US election and Brexit settled. Business will rebound and hopefully to world will recover quickly. All the best for 2021.
Way back in the 1970s, the big name in IT was IBM, and the majority of business processing took place on IBM mainframes. A lot of smaller companies were formed to produce software and hardware that provided extra functionality on those mainframes, or maybe did a similar job, but a bit faster or a bit cheaper. Two of those companies were Innovation data Processing from New Jersey, founded in 1972, and Compuware from Michigan founded in 1973. Compuware initially provided professional technical services, then expanded into mainframe software with products such as Abend-AID. Innovation is probably best known for its FDR suite of backup and recovery products.
In January 2020, Compuware formally announced that it had bought Innovation Data Processing. Both companies have a worldwide presence and a large user base. When you look at the combination of products they bring together, FDR, FDRABR, FDRCPK, FDRPAS, IAM, UPSTREAM and FATSCOPY from Innovation, and Abend-AID, Application Audit, COPE, File-AID, Hiperstation, ISPW, Strobe, ThruPut Manager, Xpediter and Topaz from Compuware, then this adds up to an impressive suite of mainframe products.
A couple of quotes from the CEOs of the two companies
Compuware CEO Chris O’Malley. "Given our mission to modernize every aspect of mainframe software delivery, the inclusion of INNOVATION’s talent, expertise, and leading backup and data protection solutions is a natural extension to our company’s culture and portfolio of leading mainframe software. Like Compuware, INNOVATION has a rich heritage of innovation that empowers customers to leverage their mainframes with ease and effectiveness. With the importance of data and compliance mandates growing, our customers will now have access to a full set of proven solutions already in use by hundreds of large enterprises worldwide."
INNOVATION’s CEO and Founder Anthony Mazzone. "Compuware is the perfect home for INNOVATION’s products and customers. Compuware has demonstrated a unique understanding of the new challenges mainframe owners face and I believe that this acquisition will best serve the needs of the mainframe storage management worldwide customer base."
For a long time, the bottleneck in computer processing has always the time it takes to send out a request to a storage system system, then wait for the data to be returned. For example, around the year 2000 a Pentium III processor ran at about 1GHz and so could run about 1,000,000,000 instructions per second, while spinning disk times were about 10 milli seconds at best, so they could return 10,000 I/Os per second. In equivalent human terms, this meant that a CPU could therefore issue an I/O request, then go out for a takeway coffee and sit and read the paper for a while before sauntering back to the office to see if the I/O had come back or not.
In practical terms, this meant that the CPU could time slice and appear to be running lots of applications simultaneously.
Moore's law said that CPU performance would double every 2 years or less, while improvements in storage disk performance lagged sone way behind. When Flash storage first arrived, it was still served by SAS protocol cabling which limited the performance improvement. Then NVMe was introduced, at first internally, then externally with NVMe-oF. The overall effect was a reduction in reponse times to about 10 microseconds, or 1000 times faster than in 2000. The latest Intel processor, the Intel core I9, runs at 5 GHz, a factor 5 increase, so Moore's is beginning to slow down and recently, storage performance has improved much faster than CPU performance. So has the performance bottleneck moved from storage to CPU?
Well, CPU instruction times are measured in nano-seconds, so they are still some way ahead of storage. The issue is the storage software. The spare CPU cycles in the storage subsystems that was available while waiting for disk I/O could be used for loads of other storage enhancements, like deduplication, compression, replication, snapshots, clones, storage tiering, disk scrubbing and error detection and correction. And there was so much spare CPU processing available that there was no drive to make these processes CPU efficient. In fact, many of them are CPU intensive. I think we will see a drive to rewrite and re-engineer this software to make it more efficient and use fewer resources, so freeing them up to be able to serve the faster I/O