IBM i with External Storage Customer References Jana

IBM Storage Systems Unformal information about IBM i customers with external storage in Europe was collected in 2009 and updated recently. Possibly there are many more customers of whom we haven’t got information. © 2011 IBM Corporation

IBM Storage Systems Customers with IBM i and DS 8000 § Spain: 20 customers – retail, bank, industry, insurance, food, distribution, public – DS 8000 copy services and Copy Services Toolkit § France: 22 customers, including – 6 Metro Mirror/ Global Mirror for DR solution – 6 IBM i mirroring – 4 software replicatio for DR – 1 Geo mirroring © 2011 IBM Corporation

IBM Storage Systems Customers with IBM i and DS 8000 - continue § Netherlands: 14 customers –Including Global mirror for DR solution § UK: 5 customers –Including Copy services Toolkit for DR solution § Sweden: 5 customers – Including Copy services Toolkit for DR solution § Denmark: 3 customers § Hungary: 2 customers –Bank, insurance § Slovakia: 2 customers –Bank © 2011 IBM Corporation

IBM Storage Systems Customers with IBM i and DS 8000 - continue § Italy: 13 customers –Including Metor Mirro / Global Mirror, Flash. Copy, and Copy services toolkit –Bank, fashion, retail § Austria: 7 customers –Bank, food –Including Metro Mirror for DR solution § Germany: 156 customers –Including Metro Mirror for DR § Israel: 2 customers –Including Metro Mirror for DR © 2011 IBM Corporation

IBM Storage Systems Customers with IBM i and DS 8000 - continue § Greece: 7 customers –Bank, food, retail § Switzerland: 10 customers –Industry, fashion –Including Metro Mirrro for DR solution and Copy services toolkit § Russia: 2 customers –bank © 2011 IBM Corporation

IBM Storage Systems Customers with IBM i and DS 5000 § Italy: 2 customers –Including Copy services toolkit § Germany: 2 customers § Austria: one customer § Nethrelands: one customer § UK: one customer § Ireland: one customer © 2011 IBM Corporation

IBM Storage Systems Customers with IBM i and XIV § Germany: 2 customers § Norway: 2 customers § Spain: one customer –Bank § Turkey: one customer with SVC and XIV (? ) § Switzerland: a benchmark with SVC and XIV was performed © 2011 IBM Corporation

IBM Storage Systems Customers with IBM i and SVC § Germany: one customer §

IBM Storage Systems Customers with IBM i and Storwize § Netherlands: 2 customers –Software replication for DR § Italy: 2 customers –Both customers run IBM i on Blade –Software replication for DR § Germany: one customer (? ) © 2011 IBM Corporation

IBM Storage Systems Following are some customer cases on which I worked; they are not official References. I am describing them as I remember, using the data and reports which I still keep. © 2011 IBM Corporation

IBM Storage Systems Banking customer - A § The customer implemented ESS with IBM i in early 2000 th § In 2005 they did a benchmark for their IBM i systems between IBM DS 8000 with Metro Mirror and EMC SRDF. DS 8000 with Metro Mirror provided better performance than the competition, the client decided for DS 8000. © 2011 IBM Corporation

IBM Storage Systems Customer’s configuration - A 2107 A. . 2107 B I 5 595 A I 5 595 B P-S L-S A Pr Switches 2107 E 2107 F I 5 595 C I 5 595 D P C Ch-S Pr-S 2107 C Site A C-S O-S Ch S 2107 G Metro Mirror of all IBM i lpars across sites Site B L O . . S-S A-S © 2011 IBM Corporation

IBM Storage Systems Banking customer - A § A DS 8000 at the customer site: 5 frames, 128 ranks (64 drive sets), RAID-5 (possibly changed to RAID-10 in the meantime), shared extent pools, storage pool striping for defining LUNs § The client implemented Production site and DR site on distance 10 km. Metro Mirror of IBM i LPARs is used accross the sites. § The customer is happy with performance of their IBM i LPARs. © 2011 IBM Corporation

IBM Storage Systems Banking customer - B § The customer implemented DS 8000 with several IBM i LPARs. § Production IBM i system used IBM i mirroring within the DS 8000 disk space. Journals were located in system ASP § On production workload in DS 8000 he noticed long waits for IO that prolonged the batch jobs. § We investigated this performance issue © 2011 IBM Corporation

IBM Storage Systems Banking customer - B § We suggested: –Stop IBM i mirroring and implement RAID-10 –Increase the cache in DS 8000 from 64 GB to 256 GB –Implement journal caching in IBM i § For testing the suggested improvements we setup a Proof of concept in Mainz. The customer sent his workload to Mainz where we restored it; he connected from remote. § The Po. C showed significant improvements in duraion of the customer’s batch job. § After the customer performed the suggested changes the resposne time in IBM i decreased, and the duration of batch job decreased for 30%. * § Experienced disk service time on DS 8000 is about 1. 4 * I don’t have the documentation anymore, 30% is what I recall © 2011 IBM Corporation

IBM Storage Systems Banking customer - C § The customer used 540 * 15 K RPM internal disk drives in mirroring § About 27 000 IO/sec including double writes/sec for mirroring or 16 000 IO/sec without mirroring § Average disk service time was 0. 8 ms, average wait time is 0, write cache efficiency 88. 4 % § He implemented – 360 * 15 K RPM internal disk in mirroring –plus 328 * LUNs on DS 8000 with 1456 GB 15 K RPM disks in RAID-10, IOP-based adapters § Average service time on is 1. 7 ms average wait time 0. 1 © 2011 IBM Corporation

IBM Storage Systems Financial services customer - D § IBM i system ASP and IASP with the application, on internal 15 K RPM disk, Geo Mirroring to DR site § The client moved both System ASP and IASP to DS 5000 with 15 K RPM disk, IASP in RAID-5 § Disk response time in IASP on internal disk: –service time 3. 9 ms, –wait time 0. 7 ms § Disk response time in IASP on DS 5000: 2. 5 ms, wait time 0. 2 ms © 2011 IBM Corporation

IBM Storage Systems Financial services customer - D § The customer experienced that some of the jobs in Batch job lasted longer when running in IASP on DS 5000, comparing to internal disk. § On our suggestion the client installed SSD in abut 10% of IASP capacity. § PEX data collection and i. Doctor queries were used to identify hot data § The hot objects were moved to SSD § Duration of the batch jobs significantly improved – IASP service time 3. 7 ms – IASP wait time 1. 7 ms § Although teh average service time and wait time are not lower than previously, teh duration of the batch job improved because of drastically better read service time on hot objects § The customer is looking for Copy Service Toolkit with DS 5000 © 2011 IBM Corporation

IBM Storage Systems Industry customer - E § The customer has several servers: – VMWare – AIX – IBM i – VIOS § They are conencted to storage systems – DS 8000 – DS 6000 – DS 4800 § IBM i: IASP with Integrated x. Servie server disk space connected with VIOS_NPIV to DS 8000, Metro Mirror of IASP § Disk service time is 5 -6 ms © 2011 IBM Corporation