Migrating from HP EVA to 3PAR StoreServ 7400. Real Implementation Experience
Greetings to all!
In this topic, I will talk about our real process of transition from the HP EVA P6300 system to a new storage system, describe our expectations and what was received from them and what not. In the text, the process of selecting systems will be very brief, although the selection process took VERY long time. I do notprovide any IOPS data so as not to start a cockfighting debate. This is not the purpose of the article. I just describe my real experience.
We are engaged in wholesale and retail trade in tools, a federal network has been built across the country, and the Central Office in the Far East. Database - 1C 8.2 soft starters, several databases of 300-500 GB, all workers. All branches work in a single database. The average number of simultaneously working in bases is from 300 to 500 people.
In 2011, I built a storage system based on the FC 8Gb protocol, consisting of 2 Brocade 8/24 switches, HP DL servers with HBA 82Q, the HP ML2024 tape library for organizing backup data storage, and the HP EVA P6300 storage system itself for data storage, mostly 1C database.
Composition of EVA: 1 shelf, 25 SAS SFF disks 300 Gb 15K. Level 1 and 5 vRAID groups were constructed for various (in terms of importance) data.
When 1C was operating in December 2012, 1C performance degraded periodically, which stopped the work of retail stores and the business processes of the company as a whole.
An analysis conducted in December 2012 - January 2013 showed that the disk subsystem is the bottleneck in the system.
So, the bottleneck is found. There were 3 solutions: to increase the number of disks by purchasing additional shelves for EVA, switch to SSD or buy storage level one.
Transferring active (the most frequently used data) to SSDs in the current EVA was not seriously considered, since EVA did not support (and will not) data tiering, and the size of the databases, taking them into vRAID-1, made such a decision insanely expensive.
When using just the second shelf, a qualitative increase in productivity when investing quite a considerable amount (more than 1.1 million rubles per shelf with 25 300 gb 15K drives) we did not receive.
Therefore, I informed the management of the company that a cardinal solution requires the replacement of a disk array. Having received approval and the maximum amount that can be spent (“but smaller is better”), the selection process has begun.
Since we are based in the Far East, vendors who were not represented by authorized SOs and representative offices further than Novosibirsk were initially thrown back, as well as for which there were many complaints about the service. For this reason, NetApp, HDS, etc. were not considered at all. HP remained in the shortlist (despite their slurred position at that time in the storage segment), IBM with their Storwize and EMC. Mandatory conditions were:
1. The presence of data tiering or similar functions (transferring frequently used data to faster disks and vice versa)
2. Support for SSD disks
3. FC 8Gb interface
4. Expandability in the future
The following data were taken for the basic configuration - 40 SAS 15K +8 SSD 200Gb drives.
Local integrators made their first offers: EMC VNX 5300 and IBM Storwize 7100. Standard bidding and clarification of the config, conditions, etc. began.
And then going to the HP website once a month as usual, I saw a description of the new 3PAR line, positioned as a middle-cost solution. I requested information. Local integrators learned about this equipment from me :), respectively, could not tell the details. The HP website is more advertising (“more, higher, faster, but we won’t tell anyone”) than technical, at that time there was almost nothing on this product (January 2013, I remind you).
I contacted the HP Far East Office, received prospectuses, and the materials are a bit more detailed. But in the Russian office of HP, they refused to give a test to the piece of iron, because there are none in Russia and it is expensive :).
By that time, it became clear that Storwize at a price does not go into the budget, IBM is very reluctant and has moved in price for a long time.
EMC (after it was "accidentally" announced the option with 3PAR) instead of the VNX5300 offered a slightly more expensive VNX5500. It was a serious bid to win. And suddenly HP said that she was ready to offset the cost of an existing and guaranteed EVA against 3PAR, leaving EVA with us. This made the solution cheaper by 18-22 kilobax more attractive.
1 QR483A 4A HP 3PAR StoreServ 7400 2-N Storage Base
20 QR492A 4A HP M6710 300GB 6G SAS 15K 2.5in HDD
20 QR492A 0D1 4A Factory integrated
4 QR503A 4A HP M6710 200GB 6G SAS 2.5in SLC SSD
4 QR503A 0D1 4A Factory integrated
1 BC795A 43 HP 3PAR 7400 Reporting Suite LTU
1 BC795A 0D1 43 Factory integrated
1 BC773A 43 HP 3PAR 7400 OS Suite Base LTU
1 BC773A 0D1 43 Factory integrated
48 BC774A 43 HP 3PAR 7400 OS Suite Drive LTU
48 BC774A 0D1 43 Factory integrated
1 BC781A 43 HP 3PAR 7400 Virtual Copy Base LTU
1 BC781A 0D1 43 Factory integrated
48 BC782A 43 HP 3PAR 7400 Virtual Copy
Base LTU 48 BC782A 0D1 43 Factory integrated
1 BC787A 43 HP 3PAR 7400 Adaptive Opt Base LTU
1 BC787A 0D1 43 Factory integrated
48 BC788A 43 HP 3PAR 7400 Adaptive Opt Drive LTU
48 BC788A 0D1 43 Factory integrated
1 BC785A 43 HP 3PAR 7400 Dynamic Opt Base LTU
1 BC785A 0D1 43 Factory integrated
48 BC786A 43 HP 3PAR 7400 Dynamic Opt Drive LTU
48 BC786A 0D1 43 Factory integrated
1 QR490A 4A HP M6710 2.5in 2U SAS Drive Enclosure
20 QR492A 4A HP M6710 300GB 6G SAS 15K 2.5in HDD
20 QR492A 0D1 4A Factory integrated
4 QR503A 4A HP M6710 200GB 6G SAS 2.5in SLC SSD
4 QR503A 0D1 4A Factory integrated
1 HA110A3 72 HP 3y Support Plus 24 SVC
1 HA110A3 RDD 6W HP 3PAR 7400 OS Suite Base LTU Supp
1 HA110A3 RDG 6W HP 3PAR 7400 Virtual Copy Base LTU Supp
1 HA110A3 RDJ 6W HP 3PAR 7400 Adaptive Opt Base LTU Supp
1 HA110A3 RDK 6W HP 3PAR 7400 Dynamic Opt Base LTU Supp
1 HA110A3 RDQ 6W HP 3PAR 7400 Reporting Suite LTU Supp
192 HA110A3 WSF 6W HP 3PAR Internal Entitlement Purpose
40 HA110A3 WUS 6W HP 3PAR 7000 Drives under 1TB Support
8 HA110A3 WUU 6W HP 3PAR 7000 200GB SSD Support
1 HA110A3 WUW 6W HP 3PAR 7000 Drive Enclosure Support
1 HA110A3 WVA 6W HP 3PAR 7400 Base
1 HA114A1 UW HP Installation and Startup Service
1 HA114A1 5TQ UW HP Startup 3PAR 7400 2-Nd Strg Base SVC
1 HA114A1 5TV UW HP Startup 3PAR 7000 2U SAS Enclosre SVC
4 QK734A 1Y HP Premier Flex LC / LC OM4 2f 5m Cbl
3 HF383A1 4J HP CP Svc for Storage Training
1 HA124A1 UW HP Technical Installation Startup SVC
1 HA124A1 5TM UW HP Startup 3PAR 7000 Reporting Ste SVC
1 HA124A1 5TB UW HP Startup 3PAR 7PAR 7PAR 7PAR Opt SVC
1 HA124A1 5TC UW HP Startup 3PAR 7000 Adaptive Opt SVC
Model with 2 controllers and the possibility of expanding if necessary 2 more.
Additional warranty - 3 years.
Delivery time - 10 weeks from prepayment.
So, 2 boxes with fully assembled (disks stuck in slots) bundles arrived at the office. Since the
HP Installation and Startup Service and HP Technical Installation Startup SVC items were paid for in the contract, I forbade my engineers to touch the boxes. Calls have begun to the supplier, to the HP Far East Office and to the Moscow office of HP with the question “When” !!! HP-DV was surprised, but after 4 days 2 engineers sent in to set up.
They immediately said that they were doing this for the first time. They started setting up on pieces of paper sent from HP-Russia. It soon became clear (on the second day) that we still needed a virtual server on VMware to deploy the management program. Server allocated.
By the end of 3 days, experts installed the devices in the rack, connected them to the existing storage system, gave addresses, passwords, logins, connected the FC-disk to the server and left. I was a little cautious that the host sees the disk as 4 devices, but the engineers know better.
On the docks from the HP website, we split volumes, set up CPG and optimization. The real performance gain was not long in coming. Even without a type of iometer, the increase in 1C was impressive.
2 months passed, and I suddenly remembered about the line of the contract - 3 pieces of HP CP Svc for Storage Training. At one time, it was included in the price to receive an additional discount (the discount amount significantly covered the cost of the position). Through HP-DV we learned that this is the cost of training 1 person on a basic setup course. The HP website suggested the date of the next courses - July 8-10.
We prepare documents and a person flies to the "non-rubber" to obtain knowledge. On the 8th, a call from him - there are deeper courses, begin on July 11, cost money. By bank transfer, we naturally do not have time to pay. Somehow at the last moment we pay in cash, we draw up a contract for an individual (in cash from a lawyer HP does not know how to draw up a contract).
Upon arrival, the specialist is finally home, we begin to analyze what and how was configured by us earlier, we understand that everything is not “feng shui”. We remake the connection method on the servers using MPIO (that's where the 4 disks that previously confused me were from), reconfiguring CPG and setting optimization according to the canons of HP.
So, the main conclusion is that the system is almost completely satisfied with the work, good (almost excellent) tools for state monitoring, performance monitoring, etc. Reports in 1C really load faster at times, there were no complaints about stopping the service. The piece of iron performs its task perfectly! At the same time, we will reorient now the old “sold” EVA to other tasks, which cannot but rejoice. By the way, I was very pleased with RAID-5 in terms of speed (at the same time, it does not rub).
upd: Technical support is good. When we, as a result of the experiment, exceeded the reserve by 85% of the volume of disks, they immediately phoned and warned.
However:
1. Why was it impossible to configure MPIO with HP engineers? Religion did not allow or did not give normal instruction?
2. Is it really impossible to give more information about the same courses, time, line as such?
3. HP Training Center - very flexible on contracts, terms of payment, etc. The company, having bought a product for far from 1-2 million rubles, cannot teach a specialist for 25 ty. with a delay of payment in 2-3 days? IMHO, nonsense. As a result, we were not able to complete the training correctly “in white”.
Of course, it must be borne in mind that this was the first experience in selling and setting up this system in the Far East, but there was "sediment".
In this topic, I will talk about our real process of transition from the HP EVA P6300 system to a new storage system, describe our expectations and what was received from them and what not. In the text, the process of selecting systems will be very brief, although the selection process took VERY long time. I do not
Company Brief
We are engaged in wholesale and retail trade in tools, a federal network has been built across the country, and the Central Office in the Far East. Database - 1C 8.2 soft starters, several databases of 300-500 GB, all workers. All branches work in a single database. The average number of simultaneously working in bases is from 300 to 500 people.
What happened
In 2011, I built a storage system based on the FC 8Gb protocol, consisting of 2 Brocade 8/24 switches, HP DL servers with HBA 82Q, the HP ML2024 tape library for organizing backup data storage, and the HP EVA P6300 storage system itself for data storage, mostly 1C database.
Composition of EVA: 1 shelf, 25 SAS SFF disks 300 Gb 15K. Level 1 and 5 vRAID groups were constructed for various (in terms of importance) data.
When 1C was operating in December 2012, 1C performance degraded periodically, which stopped the work of retail stores and the business processes of the company as a whole.
An analysis conducted in December 2012 - January 2013 showed that the disk subsystem is the bottleneck in the system.
Solution options
So, the bottleneck is found. There were 3 solutions: to increase the number of disks by purchasing additional shelves for EVA, switch to SSD or buy storage level one.
Transferring active (the most frequently used data) to SSDs in the current EVA was not seriously considered, since EVA did not support (and will not) data tiering, and the size of the databases, taking them into vRAID-1, made such a decision insanely expensive.
When using just the second shelf, a qualitative increase in productivity when investing quite a considerable amount (more than 1.1 million rubles per shelf with 25 300 gb 15K drives) we did not receive.
Therefore, I informed the management of the company that a cardinal solution requires the replacement of a disk array. Having received approval and the maximum amount that can be spent (“but smaller is better”), the selection process has begun.
Decision selection (briefly)
Since we are based in the Far East, vendors who were not represented by authorized SOs and representative offices further than Novosibirsk were initially thrown back, as well as for which there were many complaints about the service. For this reason, NetApp, HDS, etc. were not considered at all. HP remained in the shortlist (despite their slurred position at that time in the storage segment), IBM with their Storwize and EMC. Mandatory conditions were:
1. The presence of data tiering or similar functions (transferring frequently used data to faster disks and vice versa)
2. Support for SSD disks
3. FC 8Gb interface
4. Expandability in the future
The following data were taken for the basic configuration - 40 SAS 15K +8 SSD 200Gb drives.
Local integrators made their first offers: EMC VNX 5300 and IBM Storwize 7100. Standard bidding and clarification of the config, conditions, etc. began.
And then going to the HP website once a month as usual, I saw a description of the new 3PAR line, positioned as a middle-cost solution. I requested information. Local integrators learned about this equipment from me :), respectively, could not tell the details. The HP website is more advertising (“more, higher, faster, but we won’t tell anyone”) than technical, at that time there was almost nothing on this product (January 2013, I remind you).
I contacted the HP Far East Office, received prospectuses, and the materials are a bit more detailed. But in the Russian office of HP, they refused to give a test to the piece of iron, because there are none in Russia and it is expensive :).
By that time, it became clear that Storwize at a price does not go into the budget, IBM is very reluctant and has moved in price for a long time.
EMC (after it was "accidentally" announced the option with 3PAR) instead of the VNX5300 offered a slightly more expensive VNX5500. It was a serious bid to win. And suddenly HP said that she was ready to offset the cost of an existing and guaranteed EVA against 3PAR, leaving EVA with us. This made the solution cheaper by 18-22 kilobax more attractive.
As a result, the HP 3PAR option was selected in the following confe:
1 QR483A 4A HP 3PAR StoreServ 7400 2-N Storage Base
20 QR492A 4A HP M6710 300GB 6G SAS 15K 2.5in HDD
20 QR492A 0D1 4A Factory integrated
4 QR503A 4A HP M6710 200GB 6G SAS 2.5in SLC SSD
4 QR503A 0D1 4A Factory integrated
1 BC795A 43 HP 3PAR 7400 Reporting Suite LTU
1 BC795A 0D1 43 Factory integrated
1 BC773A 43 HP 3PAR 7400 OS Suite Base LTU
1 BC773A 0D1 43 Factory integrated
48 BC774A 43 HP 3PAR 7400 OS Suite Drive LTU
48 BC774A 0D1 43 Factory integrated
1 BC781A 43 HP 3PAR 7400 Virtual Copy Base LTU
1 BC781A 0D1 43 Factory integrated
48 BC782A 43 HP 3PAR 7400 Virtual Copy
Base LTU 48 BC782A 0D1 43 Factory integrated
1 BC787A 43 HP 3PAR 7400 Adaptive Opt Base LTU
1 BC787A 0D1 43 Factory integrated
48 BC788A 43 HP 3PAR 7400 Adaptive Opt Drive LTU
48 BC788A 0D1 43 Factory integrated
1 BC785A 43 HP 3PAR 7400 Dynamic Opt Base LTU
1 BC785A 0D1 43 Factory integrated
48 BC786A 43 HP 3PAR 7400 Dynamic Opt Drive LTU
48 BC786A 0D1 43 Factory integrated
1 QR490A 4A HP M6710 2.5in 2U SAS Drive Enclosure
20 QR492A 4A HP M6710 300GB 6G SAS 15K 2.5in HDD
20 QR492A 0D1 4A Factory integrated
4 QR503A 4A HP M6710 200GB 6G SAS 2.5in SLC SSD
4 QR503A 0D1 4A Factory integrated
1 HA110A3 72 HP 3y Support Plus 24 SVC
1 HA110A3 RDD 6W HP 3PAR 7400 OS Suite Base LTU Supp
1 HA110A3 RDG 6W HP 3PAR 7400 Virtual Copy Base LTU Supp
1 HA110A3 RDJ 6W HP 3PAR 7400 Adaptive Opt Base LTU Supp
1 HA110A3 RDK 6W HP 3PAR 7400 Dynamic Opt Base LTU Supp
1 HA110A3 RDQ 6W HP 3PAR 7400 Reporting Suite LTU Supp
192 HA110A3 WSF 6W HP 3PAR Internal Entitlement Purpose
40 HA110A3 WUS 6W HP 3PAR 7000 Drives under 1TB Support
8 HA110A3 WUU 6W HP 3PAR 7000 200GB SSD Support
1 HA110A3 WUW 6W HP 3PAR 7000 Drive Enclosure Support
1 HA110A3 WVA 6W HP 3PAR 7400 Base
1 HA114A1 UW HP Installation and Startup Service
1 HA114A1 5TQ UW HP Startup 3PAR 7400 2-Nd Strg Base SVC
1 HA114A1 5TV UW HP Startup 3PAR 7000 2U SAS Enclosre SVC
4 QK734A 1Y HP Premier Flex LC / LC OM4 2f 5m Cbl
3 HF383A1 4J HP CP Svc for Storage Training
1 HA124A1 UW HP Technical Installation Startup SVC
1 HA124A1 5TM UW HP Startup 3PAR 7000 Reporting Ste SVC
1 HA124A1 5TB UW HP Startup 3PAR 7PAR 7PAR 7PAR Opt SVC
1 HA124A1 5TC UW HP Startup 3PAR 7000 Adaptive Opt SVC
Model with 2 controllers and the possibility of expanding if necessary 2 more.
Additional warranty - 3 years.
Delivery time - 10 weeks from prepayment.
11 weeks passed
So, 2 boxes with fully assembled (disks stuck in slots) bundles arrived at the office. Since the
HP Installation and Startup Service and HP Technical Installation Startup SVC items were paid for in the contract, I forbade my engineers to touch the boxes. Calls have begun to the supplier, to the HP Far East Office and to the Moscow office of HP with the question “When” !!! HP-DV was surprised, but after 4 days 2 engineers sent in to set up.
They immediately said that they were doing this for the first time. They started setting up on pieces of paper sent from HP-Russia. It soon became clear (on the second day) that we still needed a virtual server on VMware to deploy the management program. Server allocated.
By the end of 3 days, experts installed the devices in the rack, connected them to the existing storage system, gave addresses, passwords, logins, connected the FC-disk to the server and left. I was a little cautious that the host sees the disk as 4 devices, but the engineers know better.
On the docks from the HP website, we split volumes, set up CPG and optimization. The real performance gain was not long in coming. Even without a type of iometer, the increase in 1C was impressive.
Error handling
2 months passed, and I suddenly remembered about the line of the contract - 3 pieces of HP CP Svc for Storage Training. At one time, it was included in the price to receive an additional discount (the discount amount significantly covered the cost of the position). Through HP-DV we learned that this is the cost of training 1 person on a basic setup course. The HP website suggested the date of the next courses - July 8-10.
We prepare documents and a person flies to the "non-rubber" to obtain knowledge. On the 8th, a call from him - there are deeper courses, begin on July 11, cost money. By bank transfer, we naturally do not have time to pay. Somehow at the last moment we pay in cash, we draw up a contract for an individual (in cash from a lawyer HP does not know how to draw up a contract).
Upon arrival, the specialist is finally home, we begin to analyze what and how was configured by us earlier, we understand that everything is not “feng shui”. We remake the connection method on the servers using MPIO (that's where the 4 disks that previously confused me were from), reconfiguring CPG and setting optimization according to the canons of HP.
Conclusion
So, the main conclusion is that the system is almost completely satisfied with the work, good (almost excellent) tools for state monitoring, performance monitoring, etc. Reports in 1C really load faster at times, there were no complaints about stopping the service. The piece of iron performs its task perfectly! At the same time, we will reorient now the old “sold” EVA to other tasks, which cannot but rejoice. By the way, I was very pleased with RAID-5 in terms of speed (at the same time, it does not rub).
upd: Technical support is good. When we, as a result of the experiment, exceeded the reserve by 85% of the volume of disks, they immediately phoned and warned.
However:
1. Why was it impossible to configure MPIO with HP engineers? Religion did not allow or did not give normal instruction?
2. Is it really impossible to give more information about the same courses, time, line as such?
3. HP Training Center - very flexible on contracts, terms of payment, etc. The company, having bought a product for far from 1-2 million rubles, cannot teach a specialist for 25 ty. with a delay of payment in 2-3 days? IMHO, nonsense. As a result, we were not able to complete the training correctly “in white”.
Of course, it must be borne in mind that this was the first experience in selling and setting up this system in the Far East, but there was "sediment".